1 API Reference

1.1 Introduction

The A2A-T SDK exposes only two entry points: the client entry A2ATClient (prompt generation, negotiation message generation and validation) and the server entry A2ATServer (prompt validation, negotiation message generation and validation). The API category overview is below.

API overview

Category	API definition	Client/Server	Meaning	Involves an LLM
Negotiation-T	`generate_negotiation_propose_prompt_from_text`	A2A-T Client / A2A-T Server	Generates a negotiation propose message from natural-language text	Yes
Negotiation-T	`generate_negotiation_accept_prompt_from_text`	A2A-T Client / A2A-T Server	Generates a negotiation accept message from natural-language text	Yes
Negotiation-T	`generate_negotiation_reject_prompt_from_text`	A2A-T Client / A2A-T Server	Generates a negotiation reject message from natural-language text	Yes
Negotiation-T	`generate_negotiation_abort_prompt_from_text`	A2A-T Client / A2A-T Server	Generates a negotiation abort message from natural-language text	Yes
Negotiation-T	`generate_negotiation_propose_prompt_from_data`	A2A-T Client / A2A-T Server	Deterministically generates a negotiation propose message from structured data	No
Negotiation-T	`generate_negotiation_accept_prompt_from_data`	A2A-T Client / A2A-T Server	Deterministically generates a negotiation accept message from structured data	No
Negotiation-T	`generate_negotiation_reject_prompt_from_data`	A2A-T Client / A2A-T Server	Deterministically generates a negotiation reject message from structured data	No
Negotiation-T	`generate_negotiation_abort_prompt_from_data`	A2A-T Client / A2A-T Server	Deterministically generates a negotiation abort message from structured data	No
Negotiation-T	`validate_propose_prompt_and_data_filling`	A2A-T Client / A2A-T Server	Validates a negotiation propose message for compliance and extracts parameters per a Schema	Yes
Negotiation-T	`validate_accept_prompt_and_data_filling`	A2A-T Client / A2A-T Server	Validates a negotiation accept message for compliance and extracts parameters per a Schema	Yes
Negotiation-T	`validate_reject_prompt_and_data_filling`	A2A-T Client / A2A-T Server	Validates a negotiation reject message for compliance and extracts parameters per a Schema	Yes
Negotiation-T	`validate_abort_prompt_and_data_filling`	A2A-T Client / A2A-T Server	Validates a negotiation abort message for compliance and extracts parameters per a Schema	Yes
Task-T	`generate_task_prompt_from_text`	A2A-T Client	Generates a task prompt from natural-language text with a specified Task-T template (skipping scenario recognition)	Yes
Task-T	`generate_task_prompt_from_data_with_schema`	A2A-T Client	Generates a task prompt from structured data + a semantic Schema with a specified Task-T template	Yes
Task-T	`validate_task_prompt_and_data_filling`	A2A-T Server	Validates a Task-T task prompt for compliance and extracts parameters per a Schema	Yes
Notification-T	`generate_notification_prompt_from_text`	A2A-T Client	Generates a notification subscription prompt from natural-language text with a specified Notification-T template	Yes
Notification-T	`generate_notification_prompt_from_data_with_schema`	A2A-T Client	Generates a notification subscription prompt from structured data + a semantic Schema with a specified Notification-T template	Yes
Notification-T	`validate_notification_prompt_and_data_filling`	A2A-T Server	Validates a Notification-T prompt for compliance and extracts parameters per a Schema	Yes
Authorization-T	`generate_auth_prompt_from_text`	A2A-T Client	Generates an authorization prompt from natural-language text with a specified Authorization-T template	Yes
Authorization-T	`generate_auth_prompt_from_data_with_schema`	A2A-T Client	Generates an authorization prompt from structured data + a semantic Schema with a specified Authorization-T template	Yes
Authorization-T	`validate_auth_prompt_and_data_filling`	A2A-T Server	Validates an Authorization-T prompt for compliance and extracts parameters per a Schema	Yes
General APIs	`generate_task_prompt`	A2A-T Client	Generates a task prompt from natural-language or structured input through scenario recognition	Yes
General APIs	`check_task_prompt`	A2A-T Server	Validates a task prompt for scenario, template, and slot compliance	Yes

Common data types and conventions

Negotiation session context: NegotiationContext(id, round, max_rounds, performative) (a frozen dataclass; id is UUID-shaped, round starts at 1, default budget DEFAULT_MAX_ROUNDS = 5), travels in the message metadata without involving the LLM; NegotiationContext.of(id, round, performative) uses the default budget, next_round() advances the round, with_performative() derives a context carrying another communicative intent, and is_exhausted() reports whether the budget is exceeded. NegotiationPerformative takes the values PROPOSE / ACCEPT / REJECT / ABORT.
Exception system: every SDK processing failure is a subclass of A2ATError; catching A2ATError covers all processing failures, and code_str gives the machine-readable error code. All messages are rendered by the SDK from error-code templates, with the language following A2AT_LANGUAGE. The following business exceptions all extend A2ATBusinessError, which additionally provides facts returning the structured fact values the rendered message is based on;
- Template generation failures raise PromptGenerationError (Task-T / Notification-T / Authorization-T) or NegotiationGenerationError (Negotiation-T).
- Template validation + parameter extraction failures raise ContentValidationError (Task-T / Notification-T / Authorization-T) or NegotiationParamExtractionError (Negotiation-T).
- Infrastructure failures are ResourceNotFoundError and ConfigFileNotFoundError, which extend A2ATError directly; programming errors (None, blank, or malformed arguments) deliberately stay outside the tree and are raised as TypeError / ValueError.

SlotValidationError: per-slot validation error details, returned with validation-failure exceptions (errors / failed_parameters in the output descriptions of the respective APIs refer to this definition):

Field	Type	Description
slot_name	str	Name of the slot that failed
code	str	Slot-level error code from the 1.4 error code list, such as `slot.not_provided`, `content.param_missing`, `content.entry_field_missing`, `content.format_error`
message	str	Human-readable error description rendered by the SDK from the code's message template; the language follows `A2AT_LANGUAGE`
facts	dict[str, str] \| None	Structured fact values the message is based on (such as `section_label`, `index`, `field_label`); may be None

Interfaces that report results through return values: generate_task_prompt and check_task_prompt do not raise business exceptions; failure information is given with the returned object (structures in 1.3.22 / 1.3.23).

Template URIs: every method the A2ATClient and A2ATServer expose that addresses a template declares the template selection through template_uri, accepts raw strings, and recommends passing the *_URI string constants of a2a_t.core.standard_templates directly (each has a same-named TemplateUri-typed constant without the _URI suffix); strings from external sources can be passed in as-is. The TemplateUri type is still used by the underlying service layers and can be parsed with TemplateUri.parse(str), returning TemplateUri | None without raising. The currently supported template URI constants (string form) are:

Constant name	Meaning	TemplateUri
standard_templates.ENERGY_SAVING_URI	Task-T energy-saving task template	Task-T/network-layer/ran-energy-saving/v1
standard_templates.PRIVATE_LINE_COMPLAINT_URI	Task-T private-line-complaint task template	Task-T/network-layer/private-line-complaint/v1
standard_templates.SUBSCRIBE_INCIDENT_URI	Notification-T incident subscription notification template	Notification-T/network-layer/subscribe-incident/v1
standard_templates.SERVICE_RECOVERY_URI	Notification-T service recovery notification template	Notification-T/network-layer/service-recovery/v1
standard_templates.AUTHORIZATION_POLICY_MANAGEMENT_URI	Authorization-T authorization policy management template	Authorization-T/authorization-policy-management/v1
standard_templates.INFORMATION_NEGOTIATION_PROPOSE_URI	Negotiation-T information negotiation propose template	Negotiation-T/information-negotiation/propose/v1
standard_templates.INFORMATION_NEGOTIATION_ACCEPT_REJECT_URI	Negotiation-T information negotiation accept/reject template	Negotiation-T/information-negotiation/accept-reject/v1
standard_templates.TARGET_NEGOTIATION_PROPOSE_URI	Negotiation-T target negotiation propose template	Negotiation-T/target-negotiation/propose/v1
standard_templates.TARGET_NEGOTIATION_ACCEPT_REJECT_URI	Negotiation-T target negotiation accept/reject template	Negotiation-T/target-negotiation/accept-reject/v1
standard_templates.FEASIBILITY_NEGOTIATION_PROPOSE_URI	Negotiation-T feasibility negotiation propose template	Negotiation-T/feasibility-negotiation/propose/v1
standard_templates.FEASIBILITY_NEGOTIATION_ACCEPT_REJECT_URI	Negotiation-T feasibility negotiation accept/reject template	Negotiation-T/feasibility-negotiation/accept-reject/v1
standard_templates.NEGOTIATION_ABORT_URI	Negotiation-T common abort template	Negotiation-T/common/abort/v1

Template URI validation policy: all methods with a template_uri parameter accept raw URI strings. A None template_uri raises TypeError; a blank or malformed URI (fewer than three segments, or a segment that is not a simple segment) raises ValueError with the message Unparseable template URI: <input>.

1.2 Constraints and Limitations

Some APIs involve LLM calls. When using them, control the call frequency and concurrency based on the concurrency the connected model service can provide and your business latency requirements.
APIs involving LLM calls apply length protection to text inputs: when the input exceeds A2AT_INPUT_TEXT_MAX_CHARS configured in .env (the package_data/.env or the file pointed to by env_path), the error code input.text_too_long is reported; the default of this configuration item is 16384 (16×1024). Structured data that does not involve LLM calls is not limited.
The SDK is stateless: negotiation session state travels with the A2A-T metadata (negotiationContext) of each message and is not stored inside the SDK; the legacy start_negotiation / receive_negotiation / continue_negotiation state-machine negotiation APIs were deprecated since v1.1.0, emitted a DeprecationWarning when called, and were removed in v1.2.0.
The private template-query methods _get_prompts / _get_prompt on A2ATClient and A2ATServer are SDK-internal interfaces. They are not part of the public API, are not guaranteed to be compatible across releases, and may be changed or removed without notice; do not call them directly.

1.3 API Descriptions

1.3.1 generate_negotiation_propose_prompt_from_text

API definition

def generate_negotiation_propose_prompt_from_text(
    self, text: str | None, context: NegotiationContext,
    template_uri: str | TemplateUri,
) -> MetadataContent

Typical scenario: after receiving a Task-T task message with incomplete parameters, the server Agent sends the client Agent a negotiation request in natural language to "supplement the missing information"; it also applies when the client Agent asks the server Agent for target clarification or a feasibility evaluation.

Function description: generates the structured negotiation message of the negotiation propose phase from natural-language text. Execution flow: load the template first, then run one LLM content extraction, and finally render the template deterministically. It applies to the initiator of information/target/feasibility negotiation.

Input description

Parameter	Type	Required	Description
text	str	Yes	Natural-language text describing the propose content (such as the list of missing information to request); the input length is limited by `A2AT_INPUT_TEXT_MAX_CHARS`, default 16384
context	NegotiationContext	Yes	Negotiation session context, injected into the `negotiationContext` metadata of the generated message without any LLM involvement
template_uri	str	Yes	Propose template, such as `INFORMATION_NEGOTIATION_PROPOSE_URI`

Request sample

from pathlib import Path
from a2a_t.client.a2at_client import A2ATClient
from a2a_t.core.metadata import MetadataContent, NegotiationContext, NegotiationPerformative
from a2a_t.core.standard_templates import INFORMATION_NEGOTIATION_PROPOSE_URI

client = A2ATClient(env_path=Path("package_data/.env"))

ctx = NegotiationContext.of(
    "3dbc13b5-bd57-4c2b-b503-24e381b6c8d3", 1, NegotiationPerformative.PROPOSE)

propose = client.generate_negotiation_propose_prompt_from_text(
    "Please provide the following missing information: Complaint category: private line "
    "interruption or poor private line quality. Both parameters are required; diagnosis "
    "cannot start without them.",
    ctx,
    INFORMATION_NEGOTIATION_PROPOSE_URI,
)

# the generated metadata travels with the A2A message
metadata = propose.build_metadata_content()

Output description

On success, returns MetadataContent:

Field/method	Type	Description
template_uri	str	Template URI used to generate the message, such as `Negotiation-T/information-negotiation/propose/v1`
prompt_text	str	Rendered negotiation message text, transferred as the value of the extension URI in the A2A message metadata
extension_uri	str	TMF extension URI (`https://projects.tmforum.org/a2aproject/telecommunication/extensions/Negotiation-T/v1`), i.e. the key of the message in the metadata
negotiation_context	NegotiationContext	Negotiation session context (id / round / maxRounds / performative), carried with the message without LLM involvement
build_metadata_content()	dict[str, object]	Builds a map that can go directly into `Message.metadata`: extension URI → message text, `templateUri` → template URI, `negotiationContext` → nested context object

On failure, raises NegotiationGenerationError (an A2ATError subclass):

Member	Type	Description
code_str	str	Machine-readable error code; values below
message	str	Human-readable failure description
facts	dict[str, str]	Structured fact values the rendered message is based on

Error codes:

template.not_found (the template or prompt resource is missing)
template.render_failed (template rendering failed)
negotiation.content_extract_failed (structured content could not be extracted from the text, retryable)
llm.not_configured (no LLM client is configured; check the A2AT_LLM_* settings)
llm.invocation_failed (LLM transport failure, retryable)
llm.response_invalid (the LLM response does not meet the step requirements, retryable)
negotiation.invalid_input (blank text, extracted content inconsistent with the phase, or a confirm request contradicting other sections)
negotiation.field_missing (a required field is missing)
input.text_too_long (the input exceeds A2AT_INPUT_TEXT_MAX_CHARS)

A None context or template_uri raises TypeError; a blank or malformed template_uri, or a performative segment that is not propose, raises ValueError; None or blank text is not a programming error and raises NegotiationGenerationError with negotiation.invalid_input (see the error codes above).

Response sample

template_uri : Negotiation-T/information-negotiation/propose/v1
extension_uri: https://projects.tmforum.org/a2aproject/telecommunication/extensions/Negotiation-T/v1
prompt_text  :
## Information Negotiation
Please supplement the relevant content based on <Required Information Items>.

## Required Information Items
1. Complaint Category: e.g. dedicated-line quality degradation

1.3.2 generate_negotiation_accept_prompt_from_text

API definition

def generate_negotiation_accept_prompt_from_text(
    self, text: str | None, context: NegotiationContext,
    template_uri: str | TemplateUri,
) -> MetadataContent

Typical scenario: after receiving the counterpart's information negotiation request, the negotiation responder (usually the client Agent) supplements/delivers the requested information in natural language and generates an accept message to return, e.g. filling in the access port name and complaint category and then confirming that diagnosis can start.

Function description: generates a negotiation accept message from natural-language text. One LLM content extraction (the extracted conclusion must be ACCEPT, otherwise rejected with negotiation.conclusion_mismatch) + deterministic rendering. It applies when the negotiation responder supplements/delivers information.

Input description

Parameter	Type	Required	Description
text	str	Yes	Natural-language text describing the accept content (such as the list of information to deliver); the input length is limited by `A2AT_INPUT_TEXT_MAX_CHARS`, default 16384
context	NegotiationContext	Yes	Negotiation session context
template_uri	str	Yes	accept-reject template, such as `INFORMATION_NEGOTIATION_ACCEPT_REJECT_URI`

Request sample

accept = client.generate_negotiation_accept_prompt_from_text(
    "Agree to supplement the following information: 1. Access Port Name: "
    "P533-Zhujiang Old Town-PTN3900-23-TPA1EG24-1; 2. Complaint Category: "
    "poor dedicated-line quality. The information is complete and diagnosis can start.",
    ctx.with_performative(NegotiationPerformative.ACCEPT),
    INFORMATION_NEGOTIATION_ACCEPT_REJECT_URI,
)

Output description

On success, returns MetadataContent (same structure as 1.3.1).

On failure, raises NegotiationGenerationError (same structure as 1.3.1). Error codes:

template.not_found (the template or prompt resource is missing)
template.render_failed (template rendering failed)
negotiation.content_extract_failed (structured content could not be extracted from the text, retryable)
llm.not_configured (no LLM client is configured; check the A2AT_LLM_* settings)
llm.invocation_failed (LLM transport failure, retryable)
llm.response_invalid (the LLM response does not meet the step requirements, retryable)
negotiation.invalid_input (blank text)
negotiation.conclusion_mismatch (the extracted conclusion is not ACCEPT)
negotiation.field_missing (a required field is missing)
input.text_too_long (the input exceeds A2AT_INPUT_TEXT_MAX_CHARS)

Response sample

template_uri : Negotiation-T/information-negotiation/accept-reject/v1
extension_uri: https://projects.tmforum.org/a2aproject/telecommunication/extensions/Negotiation-T/v1
prompt_text  :
## Information Negotiation Result
Accept

## Information Negotiation Result Content
1. Access Port Name: P533-Zhujiang Old Town-PTN3900-23-TPA1EG24-1
2. Complaint Category: dedicated-line quality degradation

1.3.3 generate_negotiation_reject_prompt_from_text

API definition

def generate_negotiation_reject_prompt_from_text(
    self, text: str | None, context: NegotiationContext,
    template_uri: str | TemplateUri,
) -> MetadataContent

Typical scenario: when the negotiation responder (usually the client Agent) cannot satisfy the counterpart's negotiation request, it generates a reject message in natural language to return and ends this negotiation round, e.g. the access port name cannot be provided because the site list is unavailable.

Function description: generates a negotiation reject message from natural-language text. One LLM content extraction (the extracted conclusion must be REJECT, otherwise rejected with negotiation.conclusion_mismatch) + deterministic rendering.

Input description: same as 1.3.2, with text being the natural-language text describing the rejection reason.

Request sample

reject = client.generate_negotiation_reject_prompt_from_text(
    "Refuse to supplement the information: the Access Port Name cannot be provided because "
    "the site list is unavailable; this negotiation ends.",
    ctx.with_performative(NegotiationPerformative.REJECT),
    INFORMATION_NEGOTIATION_ACCEPT_REJECT_URI,
)

Output description

On success, returns MetadataContent (same structure as 1.3.1).

On failure, raises NegotiationGenerationError (same structure as 1.3.1). Error codes:

template.not_found (the template or prompt resource is missing)
template.render_failed (template rendering failed)
negotiation.content_extract_failed (structured content could not be extracted from the text, retryable)
llm.not_configured (no LLM client is configured; check the A2AT_LLM_* settings)
llm.invocation_failed (LLM transport failure, retryable)
llm.response_invalid (the LLM response does not meet the step requirements, retryable)
negotiation.invalid_input (blank text)
negotiation.conclusion_mismatch (the extracted conclusion is not REJECT)
negotiation.field_missing (a required field is missing)
input.text_too_long (the input exceeds A2AT_INPUT_TEXT_MAX_CHARS)

Response sample

template_uri : Negotiation-T/information-negotiation/accept-reject/v1
extension_uri: https://projects.tmforum.org/a2aproject/telecommunication/extensions/Negotiation-T/v1
prompt_text  :
## Information Negotiation Result
Reject

## Information Negotiation Result Content
1. Access Port Name: Cannot be provided; the port resource ledger on the workbench side is temporarily unavailable

1.3.4 generate_negotiation_abort_prompt_from_text

API definition

def generate_negotiation_abort_prompt_from_text(
    self, text: str | None, context: NegotiationContext,
    template_uri: str | TemplateUri,
) -> MetadataContent

Typical scenario: when the negotiation cannot continue (round budget exhausted, timeout, token budget exhausted, etc.), either negotiation participant generates an abort message in natural language to return and terminates the whole negotiation session.

Function description: generates a negotiation abort message from natural-language text. Abort messages are independent of the negotiation type: the addressed template must be the common abort template (Negotiation-T/common/abort/v1) and the content carries only the termination reason. The flow is one LLM content extraction + deterministic rendering.

Input description

Parameter	Type	Required	Description
text	str	Yes	Natural-language text describing the termination reason; the input length is limited by `A2AT_INPUT_TEXT_MAX_CHARS`, default 16384
context	NegotiationContext	Yes	Negotiation session context
template_uri	str	Yes	Common abort template: `NEGOTIATION_ABORT_URI` (`Negotiation-T/common/abort/v1`)

Request sample

abort = client.generate_negotiation_abort_prompt_from_text(
    "The negotiation round limit is reached; the missing parameters cannot be provided; "
    "this negotiation is terminated.",
    ctx.with_performative(NegotiationPerformative.ABORT),
    NEGOTIATION_ABORT_URI,
)

Output description

On success, returns MetadataContent (same structure as 1.3.1).

On failure, raises NegotiationGenerationError (same structure as 1.3.1). Error codes:

template.not_found (the template or prompt resource is missing)
template.render_failed (template rendering failed)
negotiation.content_extract_failed (structured content could not be extracted from the text, retryable)
llm.not_configured (no LLM client is configured; check the A2AT_LLM_* settings)
llm.invocation_failed (LLM transport failure, retryable)
llm.response_invalid (the LLM response does not meet the step requirements, retryable)
negotiation.invalid_input (blank text)
negotiation.field_missing (a required field is missing)
input.text_too_long (the input exceeds A2AT_INPUT_TEXT_MAX_CHARS)

Programming errors: a None context or template_uri raises TypeError; a blank or malformed template_uri, or one not addressing the common abort template, raises ValueError.

Response sample

template_uri : Negotiation-T/common/abort/v1
extension_uri: https://projects.tmforum.org/a2aproject/telecommunication/extensions/Negotiation-T/v1
prompt_text  :
## Negotiation Result
Abort

## Negotiation Termination Reason
The negotiation round limit is reached; the missing parameters cannot be provided

1.3.5 generate_negotiation_propose_prompt_from_data

API definition

def generate_negotiation_propose_prompt_from_data(
    self, data: NegotiationProposeData, template_uri: str | TemplateUri,
) -> MetadataContent

Typical scenario: the same initiation scenario as fromText, but with structured data built by the business system as input (e.g. the server Agent auto-generating negotiation request items from the missing-slot list detected by validate_task_prompt_and_data_filling), suitable for scenarios that require deterministic message content and do not want the uncertainty of LLM extraction.

Function description: deterministically generates a negotiation propose message from structured data input, without calling the LLM. The typed content is validated, dispatched to the generator of the negotiation type addressed by the template URI, and rendered deterministically from the template.

Input description

Parameter	Type	Required	Description
data	`NegotiationProposeData(context, content)`	Yes	Negotiation context + typed propose content; the content type is selected by negotiation type
template_uri	str	Yes	Propose template

The propose content of the three negotiation types (a2a_t.negotiation.content.models):

Negotiation type	Propose content type	Fields
Information negotiation	`InformationProposeContent`	`items` (the list of missing items), `relationship` (the relationship among the missing items, nullable)
Target negotiation	`TargetProposeContent`	`target_negotiation_description` (required), `intent_understanding`, `alignment_and_clarification`, `request_for_clarification` (the three item lists can all be empty/NULL; empty omits the corresponding section), `target_confirm_request` (nullable str; non-empty means this round's message category is "target clarified and requesting confirmation from the counterpart"; when non-empty, `intent_understanding` / `alignment_and_clarification` / `request_for_clarification` must all be empty)
Feasibility negotiation	`FeasibilityProposeContent`	`feasibility_negotiation_description` (required), `action` (`NegotiationAction.REQUEST_FEASIBILITY_EVALUATION` / `PROPOSE_ALTERNATIVE_ON_FAILURE`, fixed two values), `contents_to_evaluate`, `infeasibility_details_and_proposal`, `feasibility_confirm_request` (nullable str; non-empty means this round's message category is "assessed as feasible and requesting confirmation": `action` must be `REQUEST_FEASIBILITY_EVALUATION` and both `contents_to_evaluate` / `infeasibility_details_and_proposal` must be empty)

Request sample

from a2a_t.negotiation.content.models import (
    InformationProposeContent, NegotiationItem, NegotiationProposeData,
)

ctx = NegotiationContext.of(
    "3dbc13b5-bd57-4c2b-b503-24e381b6c8d3", 2, NegotiationPerformative.PROPOSE)

propose = client.generate_negotiation_propose_prompt_from_data(
    NegotiationProposeData(
        context=ctx,
        content=InformationProposeContent(
            items=[
                NegotiationItem("Access Port Name", "e.g. P533-Zhujiang Old Town-PTN3900-23-TPA1EG24-1"),
                NegotiationItem("Complaint Category", "e.g. dedicated-line quality degradation"),
                NegotiationItem("Private Line Service Identifier", None),
            ],
            relationship="OR",
        ),
    ),
    INFORMATION_NEGOTIATION_PROPOSE_URI,
)

Output description

On success, returns MetadataContent (same structure as 1.3.1; negotiation messages all carry negotiation_context).

On failure, raises NegotiationGenerationError (same structure as 1.3.1). Error codes:

template.not_found (the template is missing)
negotiation.content_invalid (a typed content field is invalid, e.g. the required items list is empty or a required description is blank)
negotiation.invalid_input (invalid input scenarios such as a confirm request contradicting other conditional sections)
template.render_failed (template rendering failed)

There are also two kinds of programming errors (outside the A2ATError tree, standard Python exceptions): None input or its context raises TypeError; a blank or malformed template_uri, a content type inconsistent with the template's negotiation type, or a performative segment that is not propose raises ValueError.

Response sample

template_uri : Negotiation-T/information-negotiation/propose/v1
extension_uri: https://projects.tmforum.org/a2aproject/telecommunication/extensions/Negotiation-T/v1
prompt_text  :
## Information Negotiation
Please supplement the relevant content based on <Required Information Items>.

## Required Information Items
1. Access Port Name: e.g. P533-Zhujiang Old Town-PTN3900-23-TPA1EG24-1
2. Complaint Category: e.g. dedicated-line quality degradation
3. Private Line Service Identifier
Relationship between missing items: OR

1.3.6 generate_negotiation_accept_prompt_from_data

API definition

def generate_negotiation_accept_prompt_from_data(
    self, data: NegotiationEndingData, template_uri: str | TemplateUri,
) -> MetadataContent

Typical scenario: the negotiation responder (usually the client Agent) programmatically fills the parameters per the slot list requested by the counterpart and generates an accept message with structured items to return.

Function description: deterministically generates a negotiation accept message from structured data input, without calling the LLM. content.conclusion must be Accept; any other conclusion (including Abort) is rejected with the negotiation.conclusion_mismatch business error.

Input description

Parameter	Type	Required	Description
data	`NegotiationEndingData(context, content)`	Yes	Negotiation context + typed accept content (with the `Accept` conclusion)
template_uri	str	Yes	accept-reject template

The accept content of the three negotiation types: InformationEndingContent(ACCEPT, items) (the list of delivered information items), TargetEndingContent(ACCEPT, confirmed_intent, None) (the finally confirmed intent), FeasibilityEndingContent(ACCEPT, feasibility_summary) (the feasibility evaluation conclusion summary).

Request sample

from a2a_t.negotiation.content.enums import NegotiationConclusion
from a2a_t.negotiation.content.models import InformationEndingContent, NegotiationEndingData

accept = client.generate_negotiation_accept_prompt_from_data(
    NegotiationEndingData(
        context=ctx,
        content=InformationEndingContent(
            conclusion=NegotiationConclusion.ACCEPT,
            items=[
                NegotiationItem("Access Port Name", "P533-Zhujiang Old Town-PTN3900-23-TPA1EG24-1"),
                NegotiationItem("Complaint Category", "dedicated-line quality degradation"),
            ],
        ),
    ),
    INFORMATION_NEGOTIATION_ACCEPT_REJECT_URI,
)

Output description

On success, returns MetadataContent (same structure as 1.3.1).

On failure, raises NegotiationGenerationError (same structure as 1.3.1). Error codes:

template.not_found (the template is missing)
negotiation.content_invalid (a typed content field is invalid, e.g. the required items list is empty or a required description is blank)
template.render_failed (template rendering failed)

Programming errors: None input or its context raises TypeError; a blank or malformed template_uri, an inconsistent content type, or a performative segment that is not accept-reject raises ValueError; a conclusion other than ACCEPT is rejected with the negotiation.conclusion_mismatch business error.

Response sample

template_uri : Negotiation-T/information-negotiation/accept-reject/v1
extension_uri: https://projects.tmforum.org/a2aproject/telecommunication/extensions/Negotiation-T/v1
prompt_text  :
## Information Negotiation Result
Accept

## Information Negotiation Result Content
1. Access Port Name: P533-Zhujiang Old Town-PTN3900-23-TPA1EG24-1
2. Complaint Category: dedicated-line quality degradation

1.3.7 generate_negotiation_reject_prompt_from_data

API definition

def generate_negotiation_reject_prompt_from_data(
    self, data: NegotiationEndingData, template_uri: str | TemplateUri,
) -> MetadataContent

Typical scenario: after programmatically determining that the counterpart's request cannot be satisfied, the negotiation responder generates a reject message with structured items (the items that cannot be provided and the reasons) to return.

Function description: deterministically generates a negotiation reject message from structured data input, without calling the LLM. content.conclusion must be Reject; any other conclusion is rejected with the negotiation.conclusion_mismatch business error.

Input description: same as 1.3.6, but with the REJECT conclusion. Reject content: InformationEndingContent(REJECT, items) (the items that cannot be provided and the reasons), TargetEndingContent(REJECT, None, failure_reason) (the rejection reason), FeasibilityEndingContent(REJECT, feasibility_summary) (the infeasible conclusion summary).

Request sample

reject = client.generate_negotiation_reject_prompt_from_data(
    NegotiationEndingData(
        context=ctx,
        content=InformationEndingContent(
            conclusion=NegotiationConclusion.REJECT,
            items=[
                NegotiationItem(
                    "Access Port Name",
                    "Cannot be provided; the port resource ledger on the workbench side is temporarily unavailable",
                ),
            ],
        ),
    ),
    INFORMATION_NEGOTIATION_ACCEPT_REJECT_URI,
)

Output description

On success, returns MetadataContent (same structure as 1.3.1).

On failure, raises NegotiationGenerationError (same structure as 1.3.1). Error codes:

template.not_found (the template is missing)
negotiation.content_invalid (a typed content field is invalid, e.g. the required items list is empty or a required description is blank)
template.render_failed (template rendering failed)

Programming errors: None input or its context raises TypeError; a blank or malformed template_uri, an inconsistent content type, or a performative segment that is not accept-reject raises ValueError; a conclusion other than REJECT is rejected with the negotiation.conclusion_mismatch business error.

Response sample

template_uri : Negotiation-T/information-negotiation/accept-reject/v1
extension_uri: https://projects.tmforum.org/a2aproject/telecommunication/extensions/Negotiation-T/v1
prompt_text  :
## Information Negotiation Result
Reject

## Information Negotiation Result Content
1. Access Port Name: Cannot be provided; the port resource ledger on the workbench side is temporarily unavailable

1.3.8 generate_negotiation_abort_prompt_from_data

API definition

def generate_negotiation_abort_prompt_from_data(
    self, data: NegotiationAbortData, template_uri: str | TemplateUri,
) -> MetadataContent

Typical scenario: either negotiation participant generates an abort message in a structured way and terminates a negotiation session that cannot continue.

Function description: deterministically generates a negotiation abort message from structured data input, without calling the LLM. Abort messages are type-independent: the addressed template must be the common abort template and NegotiationAbortContent carries only the termination reason (the Abort conclusion is fixed template text).

Input description

Parameter	Type	Required	Description
data	`NegotiationAbortData(context, content)`	Yes	Negotiation context + `NegotiationAbortContent(termination_reason)`
template_uri	str	Yes	Common abort template: `NEGOTIATION_ABORT_URI`

Request sample

from a2a_t.negotiation.content.models import NegotiationAbortContent, NegotiationAbortData

abort = client.generate_negotiation_abort_prompt_from_data(
    NegotiationAbortData(
        context=ctx,
        content=NegotiationAbortContent(
            termination_reason="The negotiation round limit is reached; the missing parameters cannot be provided",
        ),
    ),
    NEGOTIATION_ABORT_URI,
)

Output description

On success, returns MetadataContent (same structure as 1.3.1).

On failure, raises NegotiationGenerationError (same structure as 1.3.1). Error codes:

template.not_found (the template is missing)
negotiation.content_invalid (a typed content field is invalid, e.g. a blank termination reason)
template.render_failed (template rendering failed)

Programming errors: None input or its context raises TypeError; a blank or malformed template_uri, or one not addressing the common abort template, raises ValueError.

Response sample

template_uri : Negotiation-T/common/abort/v1
extension_uri: https://projects.tmforum.org/a2aproject/telecommunication/extensions/Negotiation-T/v1
prompt_text  :
## Negotiation Result
Abort

## Negotiation Termination Reason
The negotiation round limit is reached; the missing parameters cannot be provided

1.3.9 validate_propose_prompt_and_data_filling

API definition

def validate_propose_prompt_and_data_filling(
    self,
    prompt: str | None,
    context: NegotiationContext | None,
    schema: Mapping[str, object],
    template_uri: str | TemplateUri,
) -> FilledParamData

Typical scenario: the initiator (the server Agent) self-checks outbound messages before sending a negotiation request; or the receiver (the client Agent) validates the received negotiation request and extracts the list of slots to supplement, driving the subsequent parameter filling.

Function description: validates whether a negotiation propose message is a well-formed negotiation message, and extracts parameters from it per the caller-provided JSON Schema. Pipeline order: template URI validation → deterministic rule gate (negotiation context) → template loading → one retryable LLM semantic validation (which also extracts the parameters) → deterministic parameter merging.

Input description

Parameter	Type	Required	Description
prompt	str	Yes	The negotiation propose message text to validate (`MetadataContent.prompt_text`); the input length is limited by `A2AT_INPUT_TEXT_MAX_CHARS`, default 16384
context	NegotiationContext	Yes	The negotiation context transmitted with the message; `None` reports `negotiation.invalid_input`
schema	Mapping[str, object]	Yes	Caller-provided parameter JSON Schema declaring the parameters to extract
template_uri	str	Yes	Propose template

Request sample

schema = {
    "type": "object",
    "properties": {
        "Access Port Name": {"type": "string"},
        "Complaint Category": {"type": "string"},
    },
    "required": ["Access Port Name"],
}

# propose_prompt is the negotiation propose message text sent by the counterpart
requested = client.validate_propose_prompt_and_data_filling(
    propose_prompt, ctx, schema, INFORMATION_NEGOTIATION_PROPOSE_URI,
)

# after filtering out the context parameters (id/round/maxRounds), the rest is the list of slots to supplement
needed = {k: v for k, v in requested.data.items() if k not in ("id", "round", "maxRounds")}

Output description

On success, returns FilledParamData:

Field/method	Type	Description
data	dict[str, object] \| list[object]	Merged parameters: the negotiation context parameters (`id` / `round` / `maxRounds`, winning on key conflicts) + the parameters extracted from the message per the caller's Schema; when the caller's Schema is array-shaped, `data` is an ordered array
context	dict[str, object] \| None	The context parameters (`id` / `round` / `maxRounds`) carried only when `data` is array-shaped; for an object-shaped result the context is already merged into `data` and this field is `None`

On failure, raises NegotiationParamExtractionError (an A2ATError subclass):

Member	Type	Description
code_str	str	Machine-readable error code; values below
message	str	Human-readable failure description
errors	list[SlotValidationError]	Per-slot error details; structure in the common conventions

Error codes:

negotiation.invalid_input (the prompt is None or blank, the message is not a negotiation message, or the context is None)
negotiation.rule_violation (the negotiation context violates the rules; the nested slot error codes in errors identify the specific rule, such as negotiation.invalid_context_id, negotiation.round_exceeded)
negotiation.semantic_rejected (semantic validation rejected; the per-slot details in errors use the closed negotiation.* code set, such as negotiation.conclusion_content_mismatch, negotiation.missing_result_content, negotiation.field_inconsistency)
llm.invocation_failed / llm.response_invalid (LLM failure, retryable)
template.not_found (the validation prompt resource is missing)

Programming errors: a None schema / template_uri raises TypeError; a blank or malformed template_uri, or a mismatched performative segment, raises ValueError; a None or blank prompt is not a programming error and raises NegotiationParamExtractionError with negotiation.invalid_input (see the error codes above).

Response sample

requested.data =
{
  'Access Port Name': 'e.g. P533-Zhujiang Old Town-PTN3900-23-TPA1EG24-1',
  'Complaint Category': 'e.g. dedicated-line quality degradation',
  'id': '3dbc13b5-bd57-4c2b-b503-24e381b6c8d3',
  'round': 1,
  'maxRounds': 5
}

1.3.10 validate_accept_prompt_and_data_filling

API definition

def validate_accept_prompt_and_data_filling(
    self,
    prompt: str | None,
    context: NegotiationContext | None,
    schema: Mapping[str, object],
    template_uri: str | TemplateUri,
) -> FilledParamData

Typical scenario: the initiator (the server Agent) validates the accept message returned by the counterpart, extracts the delivered parameter values, checks them against the expected filled values, and continues task execution after confirmation.

Function description: validates a negotiation accept message and extracts parameters per a Schema (the pipeline is the same as 1.3.9, with the expected performative fixed to accept-reject).

Input description: same as 1.3.9, with prompt being the accept message text and template_uri being the accept-reject template.

Request sample

# accept_prompt is the accept message text returned by the client, accept_context its negotiation context
accept_params = server.validate_accept_prompt_and_data_filling(
    accept_prompt, accept_context, schema, INFORMATION_NEGOTIATION_ACCEPT_REJECT_URI,
)

Output description

On success, returns FilledParamData (same structure as 1.3.9); data carries the parameters extracted from the delivered content per the Schema plus the context parameters.

On failure, raises NegotiationParamExtractionError (same structure as 1.3.9). Error codes:

negotiation.invalid_input (the prompt is None or blank, the message is not an accept negotiation message, or the context is None)
negotiation.rule_violation (the negotiation context violates the rules; the nested slot error codes in errors identify the specific rule, such as negotiation.invalid_context_id, negotiation.round_exceeded)
negotiation.semantic_rejected (the conclusion is not Accept or the content does not satisfy the accept-phase constraints)
llm.invocation_failed / llm.response_invalid (LLM failure, retryable)
template.not_found (the validation prompt resource is missing)

Programming errors: a None schema / template_uri raises TypeError; a blank or malformed template_uri, or a performative segment that is not accept-reject, raises ValueError; a None or blank prompt is not a programming error and raises NegotiationParamExtractionError with negotiation.invalid_input (see the error codes above).

Response sample

accept_params.data =
{
  'Access Port Name': 'P533-Zhujiang Old Town-PTN3900-23-TPA1EG24-1',
  'Complaint Category': 'dedicated-line quality degradation',
  'id': '3dbc13b5-bd57-4c2b-b503-24e381b6c8d3',
  'round': 1,
  'maxRounds': 5
}

1.3.11 validate_reject_prompt_and_data_filling

API definition

def validate_reject_prompt_and_data_filling(
    self,
    prompt: str | None,
    context: NegotiationContext | None,
    schema: Mapping[str, object],
    template_uri: str | TemplateUri,
) -> FilledParamData

Typical scenario: the initiator (the server Agent) validates the reject message returned by the counterpart, extracts the rejection reasons, and terminates the task or escalates to manual handling accordingly.

Function description: validates a negotiation reject message and extracts parameters per a Schema (the pipeline is the same as 1.3.9, with the expected performative fixed to accept-reject).

Input description: same as 1.3.9, with prompt being the reject message text and template_uri being the accept-reject template.

Request sample

reject_params = server.validate_reject_prompt_and_data_filling(
    reject_prompt, ctx, schema, INFORMATION_NEGOTIATION_ACCEPT_REJECT_URI,
)

Output description

On success, returns FilledParamData (same structure as 1.3.9); data carries the parameters extracted from the rejection reasons per the Schema plus the context parameters.

On failure, raises NegotiationParamExtractionError (same structure as 1.3.9). Error codes:

negotiation.invalid_input (the prompt is None or blank, the message is not a reject negotiation message, or the context is None)
negotiation.rule_violation (the negotiation context violates the rules; the nested slot error codes in errors identify the specific rule, such as negotiation.invalid_context_id, negotiation.round_exceeded)
negotiation.semantic_rejected (the conclusion is not Reject or the content does not satisfy the reject-phase constraints)
llm.invocation_failed / llm.response_invalid (LLM failure, retryable)
template.not_found (the validation prompt resource is missing)

Response sample

reject_params.data =
{
  'Access Port Name': 'Cannot be provided; the port resource ledger on the workbench side is temporarily unavailable',
  'id': '3dbc13b5-bd57-4c2b-b503-24e381b6c8d3',
  'round': 1,
  'maxRounds': 5
}

1.3.12 validate_abort_prompt_and_data_filling

API definition

def validate_abort_prompt_and_data_filling(
    self,
    prompt: str | None,
    context: NegotiationContext | None,
    schema: Mapping[str, object],
    template_uri: str | TemplateUri,
) -> FilledParamData

Typical scenario: a negotiation participant validates the abort message sent by the counterpart, extracts the termination reason, and releases the session resources to end the negotiation accordingly.

Function description: validates a negotiation abort message and extracts parameters per a Schema (the pipeline is the same as 1.3.9; the template_uri must address the common abort template).

Input description: same as 1.3.9, with prompt being the abort message text and template_uri being the common abort template (NEGOTIATION_ABORT_URI).

Request sample

abort_params = server.validate_abort_prompt_and_data_filling(
    abort_prompt, ctx, schema, NEGOTIATION_ABORT_URI,
)

Output description

On success, returns FilledParamData (same structure as 1.3.9); data carries the parameters extracted from the termination reason per the Schema plus the context parameters.

On failure, raises NegotiationParamExtractionError (same structure as 1.3.9). Error codes:

negotiation.invalid_input (the prompt is None or blank, the message is not an abort negotiation message, or the context is None)
negotiation.rule_violation (the negotiation context violates the rules; the nested slot error codes in errors identify the specific rule, such as negotiation.invalid_context_id, negotiation.round_exceeded)
negotiation.semantic_rejected (the message does not satisfy the abort-phase constraints)
llm.invocation_failed / llm.response_invalid (LLM failure, retryable)
template.not_found (the validation prompt resource is missing)

Programming errors: a None schema / template_uri raises TypeError; a blank or malformed template_uri, or one not addressing the common abort template, raises ValueError; a None or blank prompt is not a programming error and raises NegotiationParamExtractionError with negotiation.invalid_input (see the error codes above).

Response sample

abort_params.data =
{
  'id': '3dbc13b5-bd57-4c2b-b503-24e381b6c8d3',
  'round': 1,
  'maxRounds': 5
}

1.3.13 generate_task_prompt_from_text

API definition

def generate_task_prompt_from_text(
    self, text: str, template_uri: str | TemplateUri,
) -> MetadataContent

Typical scenario: the client Agent converts the user's natural-language task description (such as a private-line complaint diagnosis request) into a Task-T protocol message of the specified scenario; suitable when the target template is already determined and scenario recognition should be skipped.

Function description: generates a task prompt message from natural-language text with the specified Task-T template, skipping scenario recognition (the template is specified explicitly by the caller). It runs one LLM slot-extraction step and then renders the template deterministically.

Input description

Parameter	Type	Required	Description
text	str	Yes	Natural-language task description; the input length is limited by `A2AT_INPUT_TEXT_MAX_CHARS`, default 16384
template_uri	str	Yes	Task-T template, such as `PRIVATE_LINE_COMPLAINT_URI` (`Task-T/network-layer/private-line-complaint/v1`)

Request sample

from a2a_t.core.standard_templates import PRIVATE_LINE_COMPLAINT_URI

metadata = client.generate_task_prompt_from_text(
    "Help me create a private-line complaint diagnosis task for the port "
    "P781-Zhujiang New Town-PTN7900-23-TPA1EG24-17. The customer reports poor private line "
    "quality. Starting from 8:30 a.m. on May 11, 2026, the core system access from Shenzhen "
    "to Guangzhou became very slow, with latency jumping from 12ms to 320ms, and the counter "
    "and mobile banking constantly report connection timeouts. The OSS sequence number is "
    "event-id-20260511-09013.",
    PRIVATE_LINE_COMPLAINT_URI,
)

Output description

On success, returns MetadataContent:

Field/method	Type	Description
template_uri	str	Template URI used to generate the message, such as `Task-T/network-layer/private-line-complaint/v1`
prompt_text	str	Rendered task prompt message text, transferred as the value of the extension URI in the A2A message metadata
extension_uri	str	TMF extension URI (`https://projects.tmforum.org/a2aproject/telecommunication/extensions/Task-T/v1`), i.e. the key of the message in the metadata
negotiation_context	NegotiationContext	Always None (not a negotiation message)
build_metadata_content()	dict[str, object]	Builds a two-key map that can go directly into `Message.metadata`: extension URI → message text, `templateUri` → template URI

On failure, raises PromptGenerationError (an A2ATError subclass):

Member	Type	Description
code_str	str	Machine-readable error code; values below
message	str	Human-readable failure description
failed_parameters	list[SlotValidationError]	Details of the slot validation failures (non-empty for slot-domain failure codes such as `slot.not_provided`); structure in the common conventions

Error codes:

template.not_found (the template is missing)
template.load_failed (loading a prompt resource failed)
slot.schema_not_found (the slot Schema is missing)
llm.not_configured (no LLM client is configured; check the A2AT_LLM_* settings)
llm.invocation_failed / llm.response_invalid (LLM call failure, retryable)
slot.not_provided (a required slot is not provided in the input)
slot.constraint_violated (a slot value is outside the allowed range)
slot.rule_violation (fallback code for other slot validation rule violations)
template.render_failed (template rendering failed)
input.text_too_long (the input exceeds A2AT_INPUT_TEXT_MAX_CHARS)

Programming errors: a None text or template_uri raises TypeError; a blank or malformed template_uri raises ValueError.

Response sample

template_uri : Task-T/network-layer/private-line-complaint/v1
extension_uri: https://projects.tmforum.org/a2aproject/telecommunication/extensions/Task-T/v1
prompt_text  :
## Task Type
Transport private line service complaint diagnosis

## Task Description
Based on <Task Object> and <Task Context>, perform network-side fault root cause diagnosis in the complaint scenario, achieve the complaint diagnosis goal defined in <Task Target>, and return the task processing result in the structure defined in <Expected Output>.

## Task Target
Diagnose network-side faults and return diagnostic result information such as fault root causes and repair suggestions.

## Task Object
Access Port Name: P781-Zhujiang New Town-PTN7900-23-TPA1EG24-17

## Task Context
1. Complaint category: poor private line quality
2. Problem occurrence time: 2026-05-11T08:21:46Z
3. OSS-side event sequence number: event-id-20260511-09013
4. Complaint details: Starting from 8:30 a.m. on May 11, the response latency of accessing Guangzhou from Shenzhen suddenly rose from an average of 12ms to 320ms

## Expected Output
Requirement: The complaint diagnosis task result should include the following information:
1. Diagnosis result. Allowed values: success, failure (required)
2. Diagnosis result details (required)
3. Repair suggestions (optional)
4. Fault root cause list, where each fault root cause includes fault root cause name, detailed description, repair suggestions, fault root cause point location, etc. (optional)

## Terminology Explanation
1. Private line interruption
   - Synonyms: private line interruption, service interruption, network unreachable, service down, service inaccessible
2. Poor private line quality
   - Synonyms: private line poor quality, service stutter, service access timeout, service packet loss, service high latency, service jitter, service congestion, service experience degradation, service high error rate, service optical power abnormal

1.3.14 generate_task_prompt_from_data_with_schema

API definition

def generate_task_prompt_from_data_with_schema(
    self,
    data: Mapping[str, object],
    schema: Mapping[str, object],
    template_uri: str | TemplateUri,
) -> MetadataContent

Typical scenario: the client Agent converts the structured task parameters of an upstream system (field names may differ from the template slots, with the field semantics described by the Schema) into a Task-T protocol message; suitable when the task parameters are already held in structured form.

Function description: generates a task prompt from structured data + a semantic Schema with the specified Task-T template, skipping scenario recognition. The schema describes the business meaning of each input field (description / examples / enum etc.), guiding the slot filling and value constraints; each key of data corresponds to one slot value.

Input description

Parameter	Type	Required	Description
data	Mapping[str, object]	Yes	Structured business field input; keys are business field names and values are field values
schema	Mapping[str, object]	Yes (non-empty)	Field-semantics JSON Schema describing the meaning and constraints of each field
template_uri	str	Yes	Task-T template

Request sample

data = {
    "portName": "P781-Futian Center-PTN7900-2-TPA1EG24-03",
    "complaintScenario": "poor private line quality",
    "faultStartTime": "2026-05-11T08:21:46Z",
    "ticketNo": "event-id-20260511-09013",
    "faultDetailText": "Starting from 8:30 a.m. on May 11, the response latency of accessing Guangzhou from Shenzhen suddenly rose from an average of 12ms to 320ms",
}

semantics_schema = {
    "type": "object",
    "properties": {
        "portName": {"type": "string", "description": "Business field: access port name, uniquely identifying the complained private line object"},
        "complaintScenario": {
            "type": "string",
            "description": "Business field: complaint category scenario; one of private line interruption and poor private line quality is required",
            "enum": ["private line interruption", "poor private line quality"],
        },
        "faultStartTime": {"type": "string", "description": "Business field: problem occurrence time"},
        "ticketNo": {"type": "string", "description": "Business field: complaint work order or event sequence number accepted on the OSS side"},
        "faultDetailText": {"type": "string", "description": "Business field: the user's free description of the fault symptom"},
    },
    "required": ["portName", "complaintScenario"],
}

metadata = client.generate_task_prompt_from_data_with_schema(
    data, semantics_schema, PRIVATE_LINE_COMPLAINT_URI,
)

Output description

On success, returns MetadataContent (same structure as 1.3.13).

On failure, raises PromptGenerationError (same structure as 1.3.13). Programming errors: None input raises TypeError; a blank or malformed template_uri, or an empty dict schema, raises ValueError.

Response sample (same template as 1.3.13, with slot values from the structured input; Complaint details is a truncated sample value)

template_uri : Task-T/network-layer/private-line-complaint/v1
extension_uri: https://projects.tmforum.org/a2aproject/telecommunication/extensions/Task-T/v1
prompt_text  :
## Task Type
Transport private line service complaint diagnosis

## Task Description
Based on <Task Object> and <Task Context>, perform network-side fault root cause diagnosis in the complaint scenario, achieve the complaint diagnosis goal defined in <Task Target>, and return the task processing result in the structure defined in <Expected Output>.

## Task Target
Diagnose network-side faults and return diagnostic result information such as fault root causes and repair suggestions.

## Task Object
Access Port Name: P781-Futian Center-PTN7900-2-TPA1EG24-03

## Task Context
1. Complaint category: poor private line quality
2. Problem occurrence time: 2026-05-11T08:21:46Z
3. OSS-side event sequence number: event-id-20260511-09013
4. Complaint details: Starting from 8:30 a.m. on May 11, the response latency of accessing Guangzhou from Shenzhen suddenly rose from an average of 12ms to 320ms

## Expected Output
Requirement: The complaint diagnosis task result should include the following information:
1. Diagnosis result. Allowed values: success, failure (required)
2. Diagnosis result details (required)
3. Repair suggestions (optional)
4. Fault root cause list, where each fault root cause includes fault root cause name, detailed description, repair suggestions, fault root cause point location, etc. (optional)

## Terminology Explanation
1. Private line interruption
   - Synonyms: private line interruption, service interruption, network unreachable, service down, service inaccessible
2. Poor private line quality
   - Synonyms: private line poor quality, service stutter, service access timeout, service packet loss, service high latency, service jitter, service congestion, service experience degradation, service high error rate, service optical power abnormal

1.3.15 validate_task_prompt_and_data_filling

API definition

def validate_task_prompt_and_data_filling(
    self,
    prompt: str,
    schema: Mapping[str, object],
    template_uri: str | TemplateUri,
) -> FilledParamData

Typical scenario: the entry point for parameter validation and extraction after the server Agent receives a Task-T message and before entering business execution; also the decision point of "missing-slot detection" in the negotiation flow — when required parameters are missing, the caller decides whether to initiate negotiation for parameter filling.

Function description: validates whether a Task-T task prompt message matches the template and slot constraints, and extracts parameters per the caller-provided JSON Schema. Pipeline order: input gate → rule gate → template loading → one retryable LLM semantic validation (which also extracts the parameters) → deterministic parameter merging.

Input description

Parameter	Type	Required	Description
prompt	str	Yes (non-empty)	The task prompt message text to validate (`MetadataContent.prompt_text`); the input length is limited by `A2AT_INPUT_TEXT_MAX_CHARS`, default 16384
schema	Mapping[str, object]	Yes	Caller-provided parameter JSON Schema declaring the parameters to extract/validate and the required constraints
template_uri	str	Yes	Task-T template (the prefix segment must be `Task-T`)

Request sample

# Server-side parameter schema (the keys are the server's business field names, which may differ
# from the client field names; the SDK performs the cross-field adaptation)
validation_schema = {
    "type": "object",
    "properties": {
        "accessPort": {"type": "string", "description": "Access port name, uniquely identifying the complained private line object"},
        "bizScenario": {
            "type": "string",
            "description": "Complaint category scenario, required; only private line interruption and poor private line quality are allowed",
            "enum": ["private line interruption", "poor private line quality"],
        },
        "faultTime": {"type": "string", "description": "Problem occurrence time"},
        "eventSerialNo": {"type": "string", "description": "Complaint work order or event sequence number accepted on the OSS side"},
        "faultDetail": {"type": "string", "description": "Complaint/fault symptom detail description"},
    },
    "required": ["accessPort", "bizScenario"],
}

extracted = server.validate_task_prompt_and_data_filling(
    metadata.prompt_text, validation_schema, PRIVATE_LINE_COMPLAINT_URI,
).data

Output description

On success, returns FilledParamData:

Field/method	Type	Description
data	dict[str, object] \| list[object]	Parameters extracted per the Schema; keys are the parameter names declared in the Schema; when the caller's Schema is array-shaped, `data` is an ordered array
context	dict[str, object] \| None	The context parameters carried only when `data` is array-shaped; `None` for an object-shaped result

On failure, raises ContentValidationError (an A2ATError subclass):

Member	Type	Description
code_str	str	Machine-readable error code; values below
message	str	Human-readable failure description
errors	list[SlotValidationError]	Per-slot error details (slot-level error codes such as `content.param_missing`, `content.entry_field_missing`, `content.format_error`); structure in the common conventions
params	dict[str, object]	The partially extracted parameters before rejection (slots that could not be extracted have None or missing values)

Error codes:

negotiation.invalid_input (the prompt is None or blank, the schema is None, or the templateUri prefix segment/version does not match this API)
negotiation.semantic_rejected (semantic validation rejected, including missing required parameters or illegal values; the per-slot details in errors use the content.* code set, such as content.param_missing, content.entry_field_missing, content.format_error, content.value_not_allowed)
llm.invocation_failed / llm.response_invalid (LLM failure, retryable)
template.not_found (the validation prompt resource is missing)
input.text_too_long (the prompt exceeds A2AT_INPUT_TEXT_MAX_CHARS)

Programming errors: a None template_uri raises TypeError, a blank/malformed one raises ValueError; a None or blank prompt or a None schema is not a programming error and raises ContentValidationError with negotiation.invalid_input (see the error codes above).

Response sample (faultTime and faultDetail are truncated sample values)

extracted =
{
  'accessPort': 'P781-Zhujiang New Town-PTN7900-23-TPA1EG24-17',
  'bizScenario': 'poor private line quality',
  'faultTime': '2026-05-11',
  'eventSerialNo': 'event-id-20260511-09013',
  'faultDetail': '320ms'
}

On validation rejection (a negative case with a key slot missing):

ContentValidationError: [negotiation.semantic_rejected] ...
    slot=Task Object code=content.param_missing message=... facts={'section_label': 'Task Object'}

1.3.16 generate_notification_prompt_from_text

API definition

def generate_notification_prompt_from_text(
    self, text: str, template_uri: str | TemplateUri,
) -> MetadataContent

Typical scenario: the client Agent converts a natural-language subscription requirement (such as a service recovery event subscription) into a Notification-T subscription message of the specified scenario.

Function description: generates a notification subscription prompt message from natural-language text with the specified Notification-T template. It runs one LLM slot-extraction step and then renders the template deterministically; the generation phase performs built-in slot Schema validation.

Input description

Parameter	Type	Required	Description
text	str	Yes	Natural-language subscription description (notification topic, subscription condition, notification data format, etc.); the input length is limited by `A2AT_INPUT_TEXT_MAX_CHARS`, default 16384
template_uri	str	Yes	Notification-T template, such as `SUBSCRIBE_INCIDENT_URI`, `SERVICE_RECOVERY_URI`

Request sample

from a2a_t.core.standard_templates import SERVICE_RECOVERY_URI

result = client.generate_notification_prompt_from_text(
    "I want to subscribe to the service recovery event. The notification data format is: "
    "1. Service recovery plan execution status. Allowed values: not started, ended; "
    "2. Complaint diagnosis task sequence number; 3. OSS-side event sequence number; "
    "4. Access port name; 5. Whether OMC automatic recovery is authorized. Allowed values: "
    "yes, no; 6. Service recovery plan name; 7. Service recovery plan details",
    SERVICE_RECOVERY_URI,
)

Output description

On success, returns MetadataContent (same structure as 1.3.13, with extension_uri being https://projects.tmforum.org/a2aproject/telecommunication/extensions/Notification-T/v1).

On failure, raises PromptGenerationError (same structure as 1.3.13). Programming errors: a None text or template_uri raises TypeError; a blank or malformed template_uri raises ValueError.

Response sample (rendered from the template; the actual text varies with the LLM slot-extraction result; the sample input specifies no subscription condition, so that slot stays empty)

template_uri : Notification-T/network-layer/service-recovery/v1
extension_uri: https://projects.tmforum.org/a2aproject/telecommunication/extensions/Notification-T/v1
prompt_text  :
## Subscription Description
Please complete the network-side service recovery event subscription and reporting task based on the following <Notification Topic>, <Subscribe Condition>, <Notification Data Format>, and <Expected Output> information.

## Notification Topic
Service recovery event

## Subscribe Condition

## Notification Data Format
1. Service recovery plan execution status. Allowed values: not started, ended
2. Complaint diagnosis task sequence number
3. OSS-side event sequence number
4. Access port name
5. Whether OMC automatic recovery is authorized. Allowed values: yes, no
6. Service recovery plan name
7. Service recovery plan details

## Expected Output
1. Subscription result. Allowed values: success
2. After successful subscription, report messages according to <Notification Data Format>

1.3.17 generate_notification_prompt_from_data_with_schema

API definition

def generate_notification_prompt_from_data_with_schema(
    self,
    data: Mapping[str, object],
    schema: Mapping[str, object],
    template_uri: str | TemplateUri,
) -> MetadataContent

Typical scenario: the client Agent converts the structured subscription parameters submitted by an upstream system/UI into a Notification-T subscription message; suitable when the subscription parameters are already held in structured form.

Function description: generates a notification subscription prompt from structured data + a semantic Schema with the specified Notification-T template.

Input description

Parameter	Type	Required	Description
data	Mapping[str, object]	Yes	Structured subscription input (such as the subscription condition and the notification format field list)
schema	Mapping[str, object]	Yes (non-empty)	Field-semantics JSON Schema
template_uri	str	Yes	Notification-T template

Request sample

data = {
    "condition": "Subnetwork name: xx subnetwork",
    "reportFormat": [
        {"name": "Service recovery plan execution status", "values": ["not started", "ended"], "required": True},
        {"name": "Complaint diagnosis task sequence number", "required": True},
        {"name": "OSS-side event sequence number", "required": True},
        {"name": "Access port name", "required": True},
        {"name": "Whether OMC automatic recovery is authorized", "values": ["yes", "no"], "required": True},
        {"name": "Service recovery plan name", "required": True},
        {"name": "Service recovery plan details", "required": True},
        {"name": "Service recovery plan execution end time", "required": False},
    ],
}

data_schema = {
    "type": "object",
    "properties": {
        "condition": {"type": "string", "description": "Subscription condition, optional. Description of the condition to subscribe to."},
        "reportFormat": {
            "type": "array",
            "description": "Notification data format, required. The field list describing the content to report.",
            "items": {
                "type": "object",
                "properties": {
                    "name": {"type": "string"},
                    "values": {"type": "array"},
                    "required": {"type": "boolean"},
                },
                "required": ["name"],
            },
        },
    },
    "required": ["reportFormat"],
}

result = client.generate_notification_prompt_from_data_with_schema(
    data, data_schema, SERVICE_RECOVERY_URI,
)

Output description

On success, returns MetadataContent (same structure as 1.3.13, with extension_uri being https://projects.tmforum.org/a2aproject/telecommunication/extensions/Notification-T/v1).

Response sample (same template as 1.3.16, with slot values from the structured input: Subscribe Condition is filled with condition, Notification Data Format is rendered from the reportFormat list)

template_uri : Notification-T/network-layer/service-recovery/v1
extension_uri: https://projects.tmforum.org/a2aproject/telecommunication/extensions/Notification-T/v1
prompt_text  :
## Subscription Description
Please complete the network-side service recovery event subscription and reporting task based on the following <Notification Topic>, <Subscribe Condition>, <Notification Data Format>, and <Expected Output> information.

## Notification Topic
Service recovery event

## Subscribe Condition
Subnetwork name: xx subnetwork

## Notification Data Format
1. Service recovery plan execution status. Allowed values: not started, ended (required)
2. Complaint diagnosis task sequence number (required)
3. OSS-side event sequence number (required)
4. Access port name (required)
5. Whether OMC automatic recovery is authorized. Allowed values: yes, no (required)
6. Service recovery plan name (required)
7. Service recovery plan details (required)
8. Service recovery plan execution end time (optional)

## Expected Output
1. Subscription result. Allowed values: success
2. After successful subscription, report messages according to <Notification Data Format>

1.3.18 validate_notification_prompt_and_data_filling

API definition

def validate_notification_prompt_and_data_filling(
    self,
    prompt: str,
    schema: Mapping[str, object],
    template_uri: str | TemplateUri,
) -> FilledParamData

Typical scenario: after receiving a Notification-T subscription message, the server Agent validates its compliance and extracts the subscription parameters (topic/condition/reporting format), then establishes the subscription relationship accordingly.

Function description: validates whether a Notification-T notification subscription prompt message matches the template and slot constraints, and extracts parameters (subscription topic, subscription condition, notification data format, etc.) per the caller's Schema.

Input description

Parameter	Type	Required	Description
prompt	str	Yes (non-empty)	The notification subscription prompt message text to validate; the input length is limited by `A2AT_INPUT_TEXT_MAX_CHARS`, default 16384
schema	Mapping[str, object]	Yes	Caller-provided parameter JSON Schema
template_uri	str	Yes	Notification-T template (the prefix segment must be `Notification-T`)

Request sample

validation_schema = {
    "type": "object",
    "properties": {
        "topic": {"type": "string", "description": "Subscription topic (required). Name of the event topic to subscribe to."},
        "subscriptionCondition": {
            "type": "string", "description": "Subscription condition (optional). Description of the condition to subscribe to.",
        },
        "notificationDataFormat": {
            "type": "string", "description": "Notification data format (required). Description of the notification data format to report.",
        },
    },
    "required": ["topic", "notificationDataFormat"],
}

# prompt_text is the notification subscription prompt message text generated by the client
result = server.validate_notification_prompt_and_data_filling(
    prompt_text, validation_schema, SERVICE_RECOVERY_URI,
)

Output description

On success, returns FilledParamData (same structure as 1.3.15); data holds the parameters extracted per the Schema (subscription topic, subscription condition, notification data format, etc.).

On failure, raises ContentValidationError (same structure as 1.3.15). Error codes:

negotiation.invalid_input (the prompt is None or blank, the schema is None, or the templateUri prefix segment/version does not match this API)
negotiation.semantic_rejected (required parameters are missing or have illegal values; the per-slot details in errors use the content.* code set)
llm.invocation_failed / llm.response_invalid (LLM failure, retryable)
template.not_found (the validation prompt resource is missing)
input.text_too_long (the prompt exceeds A2AT_INPUT_TEXT_MAX_CHARS)

Response sample

result.data =
{
  'topic': 'Service recovery event',
  'subscriptionCondition': 'Subnetwork name: xx subnetwork',
  'notificationDataFormat': 'Service recovery event data includes: service recovery plan execution status (not started, ended), complaint diagnosis task sequence number, OSS-side event sequence number, access port name, whether OMC automatic recovery is authorized (yes, no), service recovery plan name, service recovery plan details.'
}

1.3.19 generate_auth_prompt_from_text

API definition

def generate_auth_prompt_from_text(
    self, text: str, template_uri: str | TemplateUri,
) -> MetadataContent

Typical scenario: the client Agent converts a natural-language authorization request (adding/modifying/deleting/querying a network operation authorization policy) into an Authorization-T message.

Function description: generates an authorization policy operation prompt message from natural-language text with the specified Authorization-T template. It runs one LLM slot-extraction step and then renders the template deterministically.

Input description

Parameter	Type	Required	Description
text	str	Yes	Natural-language authorization description (operation type + the authorization policy content of the network operation); the input length is limited by `A2AT_INPUT_TEXT_MAX_CHARS`, default 16384
template_uri	str	Yes	Authorization-T template: `AUTHORIZATION_POLICY_MANAGEMENT_URI` (`Authorization-T/authorization-policy-management/v1`)

Request sample

from a2a_t.core.standard_templates import AUTHORIZATION_POLICY_MANAGEMENT_URI

result = client.generate_auth_prompt_from_text(
    "Add authorization for the campus private network: use service recovery as the handling "
    "type, perform tunnel optimization, and leave the validity period to be filled in later",
    AUTHORIZATION_POLICY_MANAGEMENT_URI,
)

Output description

On success, returns MetadataContent (same structure as 1.3.13, with extension_uri being https://projects.tmforum.org/a2aproject/telecommunication/extensions/Authorization-T/v1).

On failure, raises PromptGenerationError (same structure as 1.3.13); for example, an operation type outside "add/modify/delete/query authorization policy" is rejected with slot.constraint_violated. Programming errors: a None text or template_uri raises TypeError; a blank or malformed template_uri raises ValueError.

Response sample

template_uri : Authorization-T/authorization-policy-management/v1
extension_uri: https://projects.tmforum.org/a2aproject/telecommunication/extensions/Authorization-T/v1
prompt_text  :
## Authorization Policy Operation Type
Add authorization policy

## Authorization Policy Operation Description
Please complete the corresponding authorization operation based on <Authorization Policy Operation Type> and <Network Operation Authorization Policy List>, and return the authorization policy operation execution result in the structure defined in <Expected Output>. <Expected Output> indicates the expected return content.

## Network Operation Authorization Policy List
Campus private network, service recovery, tunnel optimization

## Expected Output
1. Authorization operation execution result. Allowed values: success, failure, partial success
2. When the authorization operation is executed successfully, return the <Network Operation Authorization Policy List> that was executed successfully
3. When the authorization operation fails or is partially successful, return a failure list containing the authorization policies and the failure reasons

1.3.20 generate_auth_prompt_from_data_with_schema

API definition

def generate_auth_prompt_from_data_with_schema(
    self,
    data: Mapping[str, object],
    schema: Mapping[str, object],
    template_uri: str | TemplateUri,
) -> MetadataContent

Typical scenario: the client Agent converts the structured authorization policy data submitted field by field from an authorization management UI/system into an Authorization-T message.

Function description: generates an authorization policy operation prompt from structured data + a semantic Schema with the specified Authorization-T template, skipping scenario recognition. The semantic constraints are the same as 1.3.14.

Input description

Parameter	Type	Required	Description
data	Mapping[str, object]	Yes	Structured authorization input (operation type, policy count, policy detail list, etc.)
schema	Mapping[str, object]	Yes (non-empty)	Field-semantics JSON Schema
template_uri	str	Yes	Authorization-T template

Request sample

data = {
    "operationType": "Add authorization policy",
    "policyCount": 2,
    "details": [
        {"businessScenario": "Campus private network", "handlingType": "Service recovery",
         "operationName": "Tunnel optimization", "validityPeriod": "Permanently valid"},
        {"businessScenario": "Medical private line", "handlingType": "Service restoration",
         "operationName": "Frequency band adjustment", "validityPeriod": "2026-06-01~2030-06-18"},
    ],
}

schema = {
    "type": "object",
    "properties": {
        "operationType": {
            "type": "string",
            "enum": ["Add authorization policy", "Modify authorization policy",
                     "Delete authorization policy", "Query authorization policy"],
        },
        "policyCount": {"type": "integer", "description": "Number of policies to add"},
        "details": {
            "type": "array",
            "description": "Policy detail list",
            "items": {
                "type": "object",
                "properties": {
                    "businessScenario": {"type": "string"},
                    "handlingType": {"type": "string"},
                    "operationName": {"type": "string"},
                    "validityPeriod": {"type": "string"},
                },
            },
        },
    },
}

result = client.generate_auth_prompt_from_data_with_schema(
    data, schema, AUTHORIZATION_POLICY_MANAGEMENT_URI,
)

Output description

On success, returns MetadataContent (same structure as 1.3.13, with extension_uri being https://projects.tmforum.org/a2aproject/telecommunication/extensions/Authorization-T/v1).

Response sample

template_uri : Authorization-T/authorization-policy-management/v1
extension_uri: https://projects.tmforum.org/a2aproject/telecommunication/extensions/Authorization-T/v1
prompt_text  :
## Authorization Policy Operation Type
Add authorization policy

## Authorization Policy Operation Description
Please complete the corresponding authorization operation based on <Authorization Policy Operation Type> and <Network Operation Authorization Policy List>, and return the authorization policy operation execution result in the structure defined in <Expected Output>. <Expected Output> indicates the expected return content.

## Network Operation Authorization Policy List
Campus private network, service recovery, tunnel optimization, permanently valid; medical private line, service restoration, frequency band adjustment, 2026-06-01~2030-06-18

## Expected Output
1. Authorization operation execution result. Allowed values: success, failure, partial success
2. When the authorization operation is executed successfully, return the <Network Operation Authorization Policy List> that was executed successfully
3. When the authorization operation fails or is partially successful, return a failure list containing the authorization policies and the failure reasons

1.3.21 validate_auth_prompt_and_data_filling

API definition

def validate_auth_prompt_and_data_filling(
    self,
    prompt: str,
    schema: Mapping[str, object],
    template_uri: str | TemplateUri,
) -> FilledParamData

Typical scenario: after receiving an Authorization-T message, the server Agent validates its compliance and extracts the operation type and policy list, then executes the authorization policy management action accordingly.

Function description: validates whether an Authorization-T authorization prompt message matches the template and slot constraints, and extracts parameters (operation type, policy list, etc.) per the caller's Schema. Differentiated validation is performed per the field requirements of each operation type: added entries must include business scenario/handling type/operation name/validity period; modified entries must include the policy identifier and the new validity period; deleted entries may be the policy identifier or condition fields.

Input description

Parameter	Type	Required	Description
prompt	str	Yes (non-empty)	The authorization prompt message text to validate; the input length is limited by `A2AT_INPUT_TEXT_MAX_CHARS`, default 16384
schema	Mapping[str, object]	Yes	Caller-provided parameter JSON Schema (including `operationType`, `policies`, etc.)
template_uri	str	Yes	Authorization-T template (the prefix segment must be `Authorization-T`)

Request sample

import json

# paramSchema declares: operation type (enum: add/modify/delete/query authorization policy),
# the policy list (array; entries include policy identifier / business scenario / handling
# type / operation name / validity period)
with open("param-schema.json", encoding="utf-8") as f:
    param_schema = json.load(f)

result = server.validate_auth_prompt_and_data_filling(
    metadata.prompt_text, param_schema, AUTHORIZATION_POLICY_MANAGEMENT_URI,
)

Output description

On success, returns FilledParamData (same structure as 1.3.15); data holds the parameters extracted per the Schema (operation type, policy list, etc.).

On failure, raises ContentValidationError (same structure as 1.3.15). Error codes:

negotiation.invalid_input (the prompt is None or blank, the schema is None, or the templateUri prefix segment/version does not match this API)
negotiation.semantic_rejected (required parameters are missing or have illegal values, such as an added entry missing required fields (content.entry_field_missing) or a malformed validity period (content.format_error))
llm.invocation_failed / llm.response_invalid (LLM failure, retryable)
template.not_found (the validation prompt resource is missing)
input.text_too_long (the prompt exceeds A2AT_INPUT_TEXT_MAX_CHARS)

Response sample

result.data =
{
  'operationType': 'Add authorization policy',
  'policies': [
    {'policyId': None, 'businessScenario': 'Campus private network', 'handlingType': 'Service recovery', 'operationName': 'Tunnel optimization', 'validityPeriod': 'Permanently valid'},
    {'policyId': None, 'businessScenario': 'Medical private line', 'handlingType': 'Service restoration', 'operationName': 'Frequency band adjustment', 'validityPeriod': '2026-06-01~2030-06-18'}
  ]
}

1.3.22 generate_task_prompt

API definition

def generate_task_prompt(self, user_input: str | dict[str, object]) -> PromptGenerationResult

Typical scenario: the scenario auto-routing entry point of the client Agent: without specifying a template, the SDK recognizes the business scenario of the user input and generates the corresponding message; suitable when the scenario set is known and a simplified integration is preferred.

Function description: the general task prompt generation entry point that locates the template automatically through scenario recognition: the natural-language or structured input first goes through LLM-based business scenario recognition, and then slot extraction and rendering are completed per the built-in template of the scenario.

Input description

Parameter	Type	Required	Description
user_input	str \| dict[str, object]	Yes	Task description: `str` (natural language) or `dict` (structured input), uniformly extracted through the LLM; when the input is a `str`, its length is limited by `A2AT_INPUT_TEXT_MAX_CHARS`, default 16384

Request sample

result = client.generate_task_prompt(
    "Generate an Incident event subscription task: the notification topic is Incident, "
    "the subscription levels are critical, medium, high, and low, and the notification "
    "data format is DataPart"
)

if result.success:
    processed_prompt = result.prompt_text  # sent as A2A message metadata
else:
    print(result.failure.code, ":", result.failure.message)

Output description

On success (success is True; no exception raised; the result is returned with the return value):

Field/method	Type	Description
success	bool	Always True
prompt_text	str	Rendered task prompt message text, sent as A2A message metadata
failure	PromptGenerationFailure	Always None

On failure (success is False; no exception raised; the failure payload is returned with the result):

Field/method	Type	Description
success	bool	Always False
prompt_text	str	Always None
failure	PromptGenerationFailure	Standardized failure payload; structure in the table below

PromptGenerationFailure structure:

Field	Type	Description
code	str	Machine-readable error code from the error code list, such as `scenario.not_matched` (scenario recognition missed), `input.text_too_long` (input length protection), `template.load_failed` (prompt resource loading failed), `template.not_found` (the template is missing), `slot.schema_not_found` (the slot Schema is missing), `slot.not_provided` (a required slot is missing), `template.render_failed` (rendering failed), `llm.invocation_failed` / `llm.response_invalid` (LLM failure)
message	str	Human-readable failure description
stage	str \| None	Stage where the failure occurred: `input` (input length protection), `scenario` (scenario recognition), `preparation` (template/slot/prompt resource loading), `generation` (generation processes such as LLM slot extraction), `validation` (slot validation), `render` (template rendering)

Both PromptGenerationResult and PromptGenerationFailure provide to_dict() for direct JSON serialization.

Programming errors: a user_input that is neither a str nor a dict raises TypeError; a blank string or an empty dict raises ValueError.

Response sample

On success (the actual text varies with the LLM slot-extraction result):

result.success = True
result.prompt_text =
## Subscription Description
Based on the following <Notification Topic>, <Subscribe Condition>, <Notification Data Format>, and <Expected Output> information, complete the network-side intelligent fault Incident subscription and reporting task.

## Notification Topic
The name of this topic is "Incident"

## Subscribe Condition
The fault levels are "critical", "medium", "high", "low"

## Notification Data Format
Report Incident data via DataPart

## Expected Output
1. Subscription result, success or failure
2. Reason for subscription failure (optional)

On failure (the input matches no built-in scenario):

result.success = False
result.failure =
PromptGenerationFailure(code='scenario.not_matched', message='The input does not match any known scenario: <reason>', stage='scenario')

1.3.23 check_task_prompt

API definition

def check_task_prompt(self, *, processed_prompt_text: str) -> PromptComplianceResult

Typical scenario: the server Agent runs protocol completeness validation on the received task message (scenario/template/slot compliance); suitable when only a pass/fail conclusion is needed without parameter extraction.

Function description: the general task prompt compliance validation entry point (server): runs scenario matching, template compliance validation, and slot validation on the processed task prompt submitted by the client. Difference from validate_task_prompt_and_data_filling: this API extracts no parameters and takes no caller Schema; it only returns the standardized pass/fail conclusion.

Input description

Parameter	Type	Required	Description
processed_prompt_text	str	Yes	The A2A-T protocol message text submitted by the client (the value keyed by the extension URI in `Message.metadata`); the input length is limited by `A2AT_INPUT_TEXT_MAX_CHARS`, default 16384

Request sample

result = server.check_task_prompt(processed_prompt_text=processed_prompt)

if result.success:
    print("prompt check passed")
else:
    print(result.failure.message)

Output description

On success (success is True; no exception raised; the result is returned with the return value):

Field/method	Type	Description
success	bool	Always True
failure	PromptComplianceFailure	Always None

On failure (success is False; no exception raised; the failure payload is returned with the result):

Field/method	Type	Description
success	bool	Always False
failure	PromptComplianceFailure	Standardized failure payload; structure in the table below

PromptComplianceFailure structure:

Field	Type	Description
code	str	Machine-readable error code from the error code list: `scenario.not_matched` (message parsing/scenario recognition failed), `template.not_found` (the template is missing), slot-domain codes such as `slot.not_provided` (a required slot is missing), `slot.constraint_violated` (a value is out of range), `slot.rule_violation` (other slot rule violations), `input.text_too_long` (input length protection)
message	str	Human-readable failure description
stage	str	Stage where the failure occurred: `input_gate` (input length protection), `prompt_parse` (scenario resolution failure), `preparation` (template/slot/prompt resource loading), `slot_extraction` (LLM slot extraction), `slot_validation` (rule/semantic slot validation)

Both PromptComplianceResult and PromptComplianceFailure provide to_dict() for direct JSON serialization.

Response sample

On success:

result.success = True

On failure:

result.success = False
result.failure =
PromptComplianceFailure(code='slot.not_provided', message='\'Task Object\' is not provided in the input.', stage='slot_validation')

1.4 Error Code List

Error code categories: BUSINESS = expected business failures the caller can act on, carried by A2ATBusinessError subclasses; INFRA = infrastructure failures, carried by plain A2ATError.

Error code	Category	Message (zh-CN)	Message (en-US)
`template.not_found`	BUSINESS	模板「{template_uri}」不存在	Template '{template_uri}' does not exist
`template.render_failed`	BUSINESS	模板「{template_uri}」渲染失败:{reason}	Failed to render template '{template_uri}': {reason}
`template.load_failed`	INFRA	模板资源「{resource_path}」读取失败	Failed to read template resource '{resource_path}'
`slot.schema_not_found`	BUSINESS	模板「{template_uri}」缺少参数定义文件(语言「{language}」)	Template '{template_uri}' is missing its slot schema (language '{language}')
`slot.not_provided`	BUSINESS	输入中未提供「{slot_label}」。	'{slot_label}' is not provided in the input.
`slot.constraint_violated`	BUSINESS	「{slot_label}」的取值「{actual}」不在允许范围内	The value of '{slot_label}' ({actual}) is not within the allowed range
`slot.semantic_conflict`	BUSINESS	「{slot_label}」的取值与参数定义冲突:{reason}	The value of '{slot_label}' conflicts with the slot definition: {reason}
`slot.fabricated_value`	BUSINESS	「{slot_label}」的取值「{actual}」是占位内容,不是有效值	The value of '{slot_label}' ({actual}) is placeholder content, not a valid value
`slot.cross_scenario_pollution`	BUSINESS	「{slot_label}」的取值混入了其他场景的内容	The value of '{slot_label}' contains content from a different scenario
`slot.insufficient_grounding`	BUSINESS	「{slot_label}」的取值缺少充分依据	The value of '{slot_label}' lacks sufficient grounding
`slot.rule_violation`	BUSINESS	「{slot_label}」的取值不符合校验规则。	The value of '{slot_label}' violates the validation rules.
`input.text_too_long`	BUSINESS	输入文本长度 {actual_length} 超过上限 {max_chars}(A2AT_INPUT_TEXT_MAX_CHARS)	Input text length {actual_length} exceeds the maximum of {max_chars} (A2AT_INPUT_TEXT_MAX_CHARS)
`content.param_missing`	BUSINESS	「{section_label}」未填写,请补充该参数的取值	'{section_label}' is empty; please provide a value
`content.entry_field_missing`	BUSINESS	「{section_label}」第 {index} 条缺少必填字段「{field_label}」	Entry {index} of '{section_label}' is missing required field '{field_label}'
`content.format_error`	BUSINESS	「{section_label}」的取值格式不符合要求:{reason}	The format of '{section_label}' is invalid: {reason}
`content.value_not_allowed`	BUSINESS	「{section_label}」的取值「{actual}」不在允许范围内	The value of '{section_label}' ({actual}) is not allowed
`content.semantic_conflict`	BUSINESS	「{section_label}」存在语义冲突:{reason}	'{section_label}' has a semantic conflict: {reason}
`content.rule_violation`	BUSINESS	「{section_label}」的取值不符合校验规则。	The value of '{section_label}' violates the validation rules.
`scenario.not_matched`	BUSINESS	输入内容无法匹配任何已知场景:{reason}	The input does not match any known scenario: {reason}
`llm.not_configured`	BUSINESS	未配置 LLM 客户端,无法执行该操作(请检查 A2AT_LLM_* 配置)	No LLM client is configured; check the A2AT_LLM_* settings
`llm.invocation_failed`	BUSINESS	LLM 调用失败(提供方 {provider}):{reason}	LLM invocation failed (provider {provider}): {reason}
`llm.response_invalid`	BUSINESS	LLM 返回内容不符合要求({step} 步骤),请重试	The LLM response is invalid (step: {step}); please retry
`negotiation.invalid_input`	BUSINESS	输入的协商内容无效:{reason}	The negotiation input is invalid: {reason}
`negotiation.invalid_context_id`	BUSINESS	协商上下文标识「{actual}」不是合法的 UUID	The negotiation context id '{actual}' is not a valid UUID
`negotiation.round_exceeded`	BUSINESS	协商轮次 {round} 已超过上限 {max_rounds}	Negotiation round {round} exceeds the maximum of {max_rounds}
`negotiation.type_mismatch`	BUSINESS	报文内容属于「{implied}」协商,与声明的模板类型「{declared}」不符	The message implies '{implied}' negotiation but the declared template type is '{declared}'
`negotiation.phase_mismatch`	BUSINESS	报文阶段与声明的模板阶段不符({implied} vs {declared})	The message phase does not match the declared template phase ({implied} vs {declared})
`negotiation.conclusion_mismatch`	BUSINESS	报文结论为「{actual}」,与该方法的预期「{expected}」不符	The message conclusion is '{actual}' but '{expected}' is expected for this method
`negotiation.content_invalid`	BUSINESS	协商内容字段「{field}」无效:{reason}	The negotiation content field '{field}' is invalid: {reason}
`negotiation.field_missing`	BUSINESS	协商报文缺少必填字段「{field}」	The negotiation message is missing required field '{field}'
`negotiation.content_extract_failed`	BUSINESS	无法从文本提取协商内容({field}):{reason}	Failed to extract negotiation content from text ({field}): {reason}
`negotiation.conclusion_content_mismatch`	BUSINESS	结论为「{conclusion}」,但「{section_label}」未表达该结论应携带的内容	The conclusion is '{conclusion}' but '{section_label}' does not state the content the conclusion requires
`negotiation.missing_result_content`	BUSINESS	「{section_label}」板块缺少结论应携带的内容	The '{section_label}' section is missing the content required by its conclusion
`negotiation.mutually_exclusive_sections`	BUSINESS	互斥板块同时出现:{sections}	Mutually exclusive sections appear together: {sections}
`negotiation.constraint_conflict`	BUSINESS	「{section_label}」与既有约束冲突:{reason}	'{section_label}' conflicts with existing constraints: {reason}
`negotiation.field_inconsistency`	BUSINESS	「{section_label}」内字段取值前后不一致:{reason}	Fields within '{section_label}' are inconsistent: {reason}
`negotiation.invalid_time_interval`	BUSINESS	「{section_label}」的时间区间不合法(开始时间不得晚于结束时间)	The time interval of '{section_label}' is invalid (start must not be later than end)
`negotiation.semantic_rejected`	BUSINESS	协商报文语义校验未通过	The negotiation message failed semantic validation
`negotiation.rule_violation`	BUSINESS	「{section_label}」不符合协商报文的校验规则。	'{section_label}' violates the negotiation message validation rules.
`infra.config_invalid`	INFRA	配置项「{key}」无效:{reason}	Invalid configuration '{key}': {reason}
`infra.resource_read_failed`	INFRA	资源「{resource_path}」读取失败	Failed to read resource '{resource_path}'
`infra.internal_error`	INFRA	SDK 内部错误,请联系维护方并提供上下文	SDK internal error; contact the maintainer with context

Navigation Menu

Repository navigation

API_Reference.md

API_Reference.md

1 API Reference

1.1 Introduction

1.2 Constraints and Limitations

1.3 API Descriptions

1.3.1 generate_negotiation_propose_prompt_from_text

1.3.2 generate_negotiation_accept_prompt_from_text

1.3.3 generate_negotiation_reject_prompt_from_text

1.3.4 generate_negotiation_abort_prompt_from_text

1.3.5 generate_negotiation_propose_prompt_from_data

1.3.6 generate_negotiation_accept_prompt_from_data

1.3.7 generate_negotiation_reject_prompt_from_data

1.3.8 generate_negotiation_abort_prompt_from_data

1.3.9 validate_propose_prompt_and_data_filling

1.3.10 validate_accept_prompt_and_data_filling

1.3.11 validate_reject_prompt_and_data_filling

1.3.12 validate_abort_prompt_and_data_filling

1.3.13 generate_task_prompt_from_text

1.3.14 generate_task_prompt_from_data_with_schema

1.3.15 validate_task_prompt_and_data_filling

1.3.16 generate_notification_prompt_from_text

1.3.17 generate_notification_prompt_from_data_with_schema

1.3.18 validate_notification_prompt_and_data_filling

1.3.19 generate_auth_prompt_from_text

1.3.20 generate_auth_prompt_from_data_with_schema

1.3.21 validate_auth_prompt_and_data_filling

1.3.22 generate_task_prompt

1.3.23 check_task_prompt

1.4 Error Code List

FilesExpand file tree

API_Reference.md

Latest commit

History

API_Reference.md

File metadata and controls

1 API Reference

1.1 Introduction

1.2 Constraints and Limitations

1.3 API Descriptions

1.3.1 generate_negotiation_propose_prompt_from_text

1.3.2 generate_negotiation_accept_prompt_from_text

1.3.3 generate_negotiation_reject_prompt_from_text

1.3.4 generate_negotiation_abort_prompt_from_text

1.3.5 generate_negotiation_propose_prompt_from_data

1.3.6 generate_negotiation_accept_prompt_from_data

1.3.7 generate_negotiation_reject_prompt_from_data

1.3.8 generate_negotiation_abort_prompt_from_data

1.3.9 validate_propose_prompt_and_data_filling

1.3.10 validate_accept_prompt_and_data_filling

1.3.11 validate_reject_prompt_and_data_filling

1.3.12 validate_abort_prompt_and_data_filling

1.3.13 generate_task_prompt_from_text

1.3.14 generate_task_prompt_from_data_with_schema

1.3.15 validate_task_prompt_and_data_filling

1.3.16 generate_notification_prompt_from_text

1.3.17 generate_notification_prompt_from_data_with_schema

1.3.18 validate_notification_prompt_and_data_filling

1.3.19 generate_auth_prompt_from_text

1.3.20 generate_auth_prompt_from_data_with_schema

1.3.21 validate_auth_prompt_and_data_filling

1.3.22 generate_task_prompt

1.3.23 check_task_prompt

1.4 Error Code List