Repository navigation
Conversation
Signed-off-by: gcw_mr3RVhmf <cnafan@outlook.com>
Upstream renamed validateAndFillingTaskData to validateTaskPromptAndDataFilling; the sample executor, strategy javadoc, scenario loader javadoc and diagnosis service references now use the new name. Signed-off-by: ZhouJie6 <ivo.zhou@huawei.com>
…n-sample feat(sample): add negotiation end-to-end demo and fromData/fromText API samples
Signed-off-by: gcw_mr3RVhmf <cnafan@outlook.com>
…ngle-brace fix(prompt): keep single-brace example text verbatim instead of failing as unknown slot
Signed-off-by: gcw_mr3RVhmf <cnafan@outlook.com>
…parameters and drop CI e2e job Split the server-side notificationDataFormat string parameter in the service-recovery sample schema into the 10 individual business fields defined by the template (7 required + 3 optional), update the required list and slot names accordingly, and align the sample inputs with the new parameter names. Remove the e2e-samples-unified job from CI. Signed-off-by: wangruiyang8 <wangruiyang8@huawei.com>
Signed-off-by: qinyanhao01 <qinyanhao1@huawei.com> # Conflicts: # a2a-t-sample/README.zh-CN.md
Signed-off-by: gcw_mr3RVhmf <cnafan@outlook.com>
* origin/main: fix(sample): adapt to validateTaskPromptAndDataFilling rename feat(sample): stream transport support and release-grade design document fix(client): use double-braced unknown slot to trigger render failure in tests fix(prompt): keep single-brace example text verbatim instead of failing as unknown slot refactor(sample): keep only generic rules in negotiation code, move all scenario values to scenario.json refactor(sample): load negotiation scenario data from scenario.json feat(sample): add negotiation end-to-end demo and fromData/fromText API samples fix(client): use double-braced unknown slot to trigger render failure in tests fix(prompt): keep single-brace example text verbatim instead of failing as unknown slot Signed-off-by: gcw_mr3RVhmf <cnafan@outlook.com> # Conflicts: # a2a-t-sample/README.zh-CN.md # a2a-t-sample/pom.xml
Signed-off-by: gcw_mr3RVhmf <cnafan@outlook.com>
# Conflicts: # a2a-t-sample/README.zh-CN.md
…egotiation-evaluation feat(sample): add private-line negotiation sample
Signed-off-by: qinyanhao01 <qinyanhao1@huawei.com>
…istency Add service-recovery-client.javaargs.txt and service-recovery-server.javaargs.txt to the pom, matching the pattern used by negotiation and other samples. Document the new startup method in README.zh-CN.md. Also fix the validateAndFillingNotificationData rename to validateNotificationPromptAndDataFilling that broke service-recovery compilation. Signed-off-by: wangruiyang8 <wangruiyang8@huawei.com>
feat(sample): add service-recovery sample with unified SampleRunner
feat(sample):add task-t client env template
Signed-off-by: qinyanhao01 <qinyanhao1@huawei.com>
Signed-off-by: qinyanhao01 <qinyanhao1@huawei.com> # Conflicts: # a2a-t-sample/pom.xml
…playable JSON report Adds NegotiationEvalApp, a pure-Java evaluation entry under the negotiation sample that drives the negotiation prompt interfaces in message order for every case of the bundled eval-suite.json: 1. Task-T generation (generateTaskPromptFromDataWithSchema / fromText) 2. server validation with missing-slot detection (validateTaskPromptAndDataFilling) 3. Negotiation-T propose generation from the detected missing slots (generateNegotiationProposePromptFromData) 4. outbound propose validation (validateProposePromptAndDataFilling) 5. client fill + accept generation (generateNegotiationAcceptPromptFromData) 6. inbound accept validation with fill-value comparison (validateAcceptPromptAndDataFilling) 7. second Task-T validation deciding COMPLETED vs INPUT_REQUIRED Every intermediate artifact (generated prompts, validation verdicts, extracted parameters, per-slot errors, client fill values) is recorded per case in a JSON report that is rewritten after each case, so a single case can be replayed and audited without re-running it. The suite carries the 10 private-line-complaint cases (fromData/fromText channels, generation fail-fast and value-constraint negatives) with expected outcomes; metrics cover trigger accuracy, missing-slot exact match, propose outbound validity, fill-value match and closed-loop completion. Run: java @a2a-t-sample/target/eval.javaargs.txt [--case PLC-04] [--out eval-report.json] /path/to/.env Signed-off-by: ZhouJie6 <ivo.zhou@huawei.com>
…ctable DefaultSemanticValidator.parseParams dropped parameters whose extracted value was null, so FilledParamData did not even carry the key. A null param is the semantic validator's explicit signal that a schema slot is missing from the content (per the content_validation user prompt: slots that cannot be extracted must be output as null). With the key gone, any caller that detects missing parameters by scanning FilledParamData for blank values cannot see the missing slot, and negotiation is silently never triggered for it. The parse now keeps null-valued keys (the normalized map is wrapped with Collections.unmodifiableMap instead of Map.copyOf, which rejects null values), and ValidationPipeline.mergeParams already carries them through unchanged. Found by the negotiation interface closed-loop evaluator (eval-suite case PLC-10: the LLM returned 任务对象=null with semantic_verdict=true, the missing slot vanished from FilledParamData, and the negotiation flow never started). Signed-off-by: ZhouJie6 <ivo.zhou@huawei.com>
… interface evaluator - add a dedicated subsection explaining how the javaargs.txt files work (JDK 9+ @-argfile syntax, generated by the antrun plugin at package time, one file per entry point) so the run commands are readable without Maven knowledge; note the Windows -Dfile.encoding=UTF-8 hint - document the eval entry: full-suite / --case replay / --out forms, incremental report writing, and where the eval suite lives - list the evaluator in the code-structure tree and the coverage matrix Signed-off-by: ZhouJie6 <ivo.zhou@huawei.com>
…n-interface-eval feat(sample): negotiation interface closed-loop evaluator with replayable JSON report
…lidator-null-params fix(prompt): preserve null semantic params so missing slots stay detectable
Relocate the negotiation session context record from the negotiation content package to core.model so MetadataContent can reference it without a reverse module dependency. The record is moved verbatim (id/round/maxRounds with validation, DEFAULT_MAX_ROUNDS, of/nextRound/ isExhausted), its test mirrors the new package, and all imports and fully-qualified references across core/negotiation/client/server/sample are updated. The deprecated types.model.NegotiationContext demo class and the negotiation package-structure pin are left in sync. Signed-off-by: codetalks-hw <lihaizhen5@huawei.com>
Add an optional negotiationContext component to MetadataContent together
with the 3-arg convenience constructor delegating a null context, so
non-negotiation messages keep their two-key metadata shape. The
negotiation generation orchestrator now passes its context through, and
buildMetadataContent() returns Map<String,Object> carrying, for
negotiation messages only, a third negotiationContext key mapped to the
nested {id, round, maxRounds} object matching the A2A-T message shape.
Contract tests now pin three keys on the negotiation path and two keys
elsewhere.
Signed-off-by: codetalks-hw <lihaizhen5@huawei.com>
The four validate*PromptAndDataFilling methods on NegotiationContentService, A2ATClient and A2ATServer now take the negotiation context as their second argument, mirroring the from-text generation APIs, and the compliance checker validates the passed context (UUID shape, round within budget) instead of re-parsing a rendered context section. NegotiationRuleCheckResult shrinks to (passed, errors), the rule-checker adapter becomes a per-call bridge carrying the context, and a null context is reported as negotiation_invalid_input. The private-line sample reads the context back from the negotiationContext metadata entry through NegotiationMetadataReader.readContext and feeds it into the validation calls, replacing the round trip through the rendered text. Signed-off-by: codetalks-hw <lihaizhen5@huawei.com>
The negotiation session context now travels exclusively in the negotiationContext metadata entry, so the rendered message no longer carries a context section: all fourteen Negotiation-T templates drop their leading context section, the seven generators stop injecting the slot.context value, NegotiationContextRenderer and its dedicated tests are deleted, the section.context/slot.context keys leave both vocabulary files together with Vocabulary.CANONICAL_KEYS, the template lint profiles and the negotiation-context rule are dropped, and the semantic-validation system prompts lose the section.context slot entry and the id example parameter. The eighteen golden fixtures are regenerated (each losing exactly the five context lines) and reviewed, the assertion suite now pins the absence of the context section alongside the metadata context, and the private-line evaluation corpus drops the hand-written context headers from the client supplement and all one hundred completed prompts. Signed-off-by: codetalks-hw <lihaizhen5@huawei.com>
The developer guides now describe the four-component MetadataContent with
the conditional negotiationContext metadata key carrying the nested
{id, round, maxRounds} object, the NegotiationContext home in the core
model package, the context-parameter signatures of the four
validate*PromptAndDataFilling methods, and the rule gate validating the
caller-supplied context instead of parsing a rendered section. The
end-to-end examples pass the context through and the method-count drift
(thirteen vs the actual fourteen facade methods) is corrected, as is the
same drift in the two facade symmetry tests that missed the abort
methods.
Signed-off-by: codetalks-hw <lihaizhen5@huawei.com>
…ationContext validate APIs Signed-off-by: codetalks-hw <lihaizhen5@huawei.com>
fix(resources): update ran-energy-saving template
* dev: Merge pull request project-openan#193 from zhangwenjun-huawei/dev fix(resources): update ran-energy-saving template Merge pull request project-openan#192 from zhangwenjun-huawei/dev feat(llm): add DEBUG-level LLM call logging with opt-in payload switch Merge pull request project-openan#190 from zhangwenjun-huawei/dev feat(corpus): add Negotiation-T accuracy verification module and unify suite directories Merge pull request project-openan#189 from mamusen-huawei/dev fix(resources): enable required slots in ran-energy-saving en-US slot.json style(corpus): normalize ran-energy-saving modify input case JSON to benchmark style test(corpus): add ran-energy-saving modify input cases with unified schema Merge pull request project-openan#187 from Zhoujie628/dev feat(llm): add configurable TLS verification Merge pull request project-openan#183 from zhangwenjun-huawei/dev feat(corpus): rewrite a2a-t-corpus as a data-driven accuracy verification module Merge pull request project-openan#182 from zhangwenjun-huawei/dev chore: add .gitattributes and normalize line endings to LF Merge pull request project-openan#181 from zhangwenjun-huawei/dev fix(ci): allow Terminology Explanation in Authorization-T lint profile ci: restrict push and pull_request triggers to main and dev branches fix(doc): refresh template.md and docs
Merge branch 'dev' into main
Signed-off-by: zhangwenjun-huawei <zhangwenjun19@huawei.com>
chore(release): bump version to 1.1.1
Signed-off-by: zhangwenjun-huawei <zhangwenjun19@huawei.com>
chore(ci): exclude corpus module from release pipeli
…mple transitive resolution
Owner
Author
|
Created in error (wrong repo). The real PR is at https://github.com/project-openan/a2a-t-sdk-java — closing this one. |
Owner
Author
|
Created in error (wrong repo). The real PR is at https://github.com/project-openan/a2a-t-sdk-java - closing this one. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
概述
按设计文档 docs/superpowers/specs/2026-10-20-observability-java-port-design.md 落地 A2A-T Java SDK 可观测性（Plan A：核心模块 + SPI + 门面接线；Python SDK 可观测模块的 Java 移植）。协议层装饰器（Plan B）另行提交。
交付物
关键设计决策
质量