fix(agent): preserve provider state and signed tool-loop prefixes by hpiclaranet · Pull Request #41095 · langgenius/dify

hpiclaranet · 2026-08-22T11:20:53Z

Summary

Part of #41092: preserve provider replay state during one Function Calling Agent execution.
The runner previously reconstructed replies from visible text and tool calls, then changed the history window, image inputs, or tool schemas before the next request.
It now carries the last non-None opaque_body snapshot in both streaming and blocking modes, chooses the initial history once, freezes tool schemas, and appends each assistant/tool turn without changing the signed prefix.
The iteration limit no longer clears tools; an exhausted context budget stops before invocation without reducing output tokens below the model's configured thinking budget.
Configured or schema-default output budgets below 16 tokens remain valid; the fallback applies only when no budget is available.

The host regression exercises the real PluginModelClient, daemon-response Pydantic decoding, RequestInvokeLLM, runtime normalization, and backwards response reconstruction for Anthropic/OpenAI payloads in both modes.
HTTP transport is mocked at the daemon boundary; this is not a deployed end-to-end API test.
The change retains the original PR author's commits and extends the same PR.

Dependencies and release gate

This remains a draft.
Dify's Graphon 0.7 runtime API is incompatible with Graphon main/0.8, so the review build pins the tested 0.7 backport in api/pyproject.toml and api/uv.lock.
The mainline fix is Graphon #261.
Replace the temporary source override with an actual containing release before merging for production.

The Agent strategy plugin additionally requires SDK #390 and daemon #828.
Provider counterparts are Anthropic #3919 and OpenAI #3918.
The final daemon image and minimum supported Dify version must refer to actual published fixes; no future release version is assumed here.
Cross-session history reconstruction and third-party provider adapters are outside this in-memory tool-loop fix.

Validation

Graphon 0.7 backport: 759 unit tests passed, plus type/format checks and independent test, knowledge, and naming reviews.
Dify with the updated immutable Graphon source installed through uv: 116 targeted tests passed across the FC runner and host transport/model boundaries.
The backport now also covers Slim ordinary/structured blocking aggregation and prompt filtering, including assistant state attached to empty content arrays.
Affected Ruff checks, formatting, complete make type-check-core (Pyrefly plus Mypy across 1868 source files), lock validation, and git diff --check passed.
After correcting the new test annotations and using a real Tenant entity, all 6 replay transport cases passed again.
Regressions cover stream/blocking snapshots, empty snapshots, image inputs, mutable nested tool schemas, history-window stability, context budgets, and iteration limits.
At the previous head a637fc4e72997948c615529565cb3a9d53a3be6b, Main CI and Docker builds passed, including API unit/controller/integration/coverage, Python Style, VDB, browser E2E, and API/Web/local-sandbox images.
One Chromium shard initially could not pull Weaviate due to Docker Hub rate limiting; a single isolated retry passed without rerunning the successful backend tests.
CI for the new dependency and small-budget revision is tracked separately from those earlier successful runs.
Live provider API requests and a deployed Dify/daemon/plugin combination have not been tested.

Screenshots

Not applicable: backend behavior only.

Checklist

This change requires a documentation update, included: Dify Document.
There is an associated issue and I have assigned it to myself.
Tests cover the changed behavior.
Targeted lint and full core type checks passed; remaining full checks are delegated to CI.

api/core/rag/extractor/unstructured/unstructured_epub_extractor.py imports pypandoc directly, but it was only satisfied transitively through graphon's PyPI wheel (unstructured[epub] extra). Declare it directly, with the bundled pypandoc-binary so the Docker image does not need a system pandoc install.

…agent iterations FunctionCallAgentRunner rebuilt each assistant turn from the flattened text plus tool calls only, so provider round-trip state exposed by plugins via AssistantPromptMessage.opaque_body (e.g. signed/encrypted Claude thinking blocks) was lost between LLM iterations, causing provider 400 errors on the next tool-use round. Capture the complete snapshot from the LLM result (streaming chunks: last non-None wins; blocking result: message.opaque_body) and attach it to the assistant message stored in _current_thoughts so the next iteration's prompt carries it back to the provider. Also point the graphon dependency at the cherry-picked opaque_body change (rev-pinned, v0.7.0-based, zero drift) until the release containing langgenius/graphon#261 is published to PyPI; the pin then becomes graphon==0.7.1. Part of langgenius#41092; plugin-side counterpart: langgenius/dify-official-plugins#3715; core entity change: langgenius/graphon#261.

hpiclaranet · 2026-08-22T11:24:02Z

Heads-up to maintainers — this draft needs two things from the core side:

Workflow approval for the fork PR (several checks are action_required).
Review of the companion PR langgenius/graphon#261 (entity/accumulator change, small and additive). This draft pins graphon at the cherry-picked commit of that PR (zero drift from the current 0.7.0 pin) so its CI is green. It is a draft because it cannot merge until the graphon release containing fix: stop response btn hide messages #261 is published — then the temporary [tool.uv.sources] graphon override comes out and the pin becomes graphon==0.7.1.

One heads-up for that eventual pin bump: graphon main has drifted since the 0.7.0 release (added odfdo, removed the unstructured[epub] extra that was transitively providing pypandoc, LLM node changes). The chore(deps) commit here already declares pypandoc/pypandoc-binary directly (they were an undeclared direct import of core/rag/extractor/unstructured), but the 0.7.1 bump will likely need a few dify test updates for the graphon main behavioral changes.

Tracking: #41092 · graphon #260/#261 · plugin side dify-official-plugins#3715.

github-actions · 2026-08-22T11:24:56Z

Pyrefly Diff

base → PR

--- /tmp/pyrefly_base.txt	2026-09-24 10:21:19.903552615 +0000
+++ /tmp/pyrefly_pr.txt	2026-09-24 10:21:10.267517606 +0000
@@ -2375,89 +2375,135 @@
 ERROR `in` is not supported between `Literal['Question: What is Python?']` and `None` [not-iterable]
    --> tests/unit_tests/core/agent/test_cot_completion_agent_runner.py:180:16
 ERROR `Literal[0]` is not assignable to attribute `prompt_price` with type `Decimal` [bad-assignment]
-  --> tests/unit_tests/core/agent/test_fc_agent_runner.py:42:26
+  --> tests/unit_tests/core/agent/test_fc_agent_runner.py:47:26
 ERROR `Literal[0]` is not assignable to attribute `completion_price` with type `Decimal` [bad-assignment]
-  --> tests/unit_tests/core/agent/test_fc_agent_runner.py:43:30
+  --> tests/unit_tests/core/agent/test_fc_agent_runner.py:48:30
 ERROR `Literal[0]` is not assignable to attribute `total_price` with type `Decimal` [bad-assignment]
-  --> tests/unit_tests/core/agent/test_fc_agent_runner.py:44:25
+  --> tests/unit_tests/core/agent/test_fc_agent_runner.py:49:25
 ERROR Missing argument `session` in function `core.agent.base_agent_runner.BaseAgentRunner.__init__` [missing-argument]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:144:37
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:150:37
 ERROR Object of class `FunctionCallAgentRunner` has no attribute `session` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:176:5
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:182:5
 ERROR Object of class `FunctionCallAgentRunner` has no attribute `session` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:187:9
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:193:9
 ERROR Argument `DummyChunk` is not assignable to parameter `llm_result_chunk` with type `LLMResultChunk` in function `core.agent.fc_agent_runner.FunctionCallAgentRunner.check_tool_calls` [bad-argument-type]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:199:40
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:205:40
 ERROR Argument `DummyResult` is not assignable to parameter `llm_result` with type `LLMResult` in function `core.agent.fc_agent_runner.FunctionCallAgentRunner.check_blocking_tool_calls` [bad-argument-type]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:204:49
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:210:49
 ERROR Argument `DummyChunk` is not assignable to parameter `llm_result_chunk` with type `LLMResultChunk` in function `core.agent.fc_agent_runner.FunctionCallAgentRunner.extract_tool_calls` [bad-argument-type]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:220:43
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:226:43
 ERROR Argument `DummyChunk` is not assignable to parameter `llm_result_chunk` with type `LLMResultChunk` in function `core.agent.fc_agent_runner.FunctionCallAgentRunner.extract_tool_calls` [bad-argument-type]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:231:43
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:237:43
 ERROR Argument `DummyResult` is not assignable to parameter `llm_result` with type `LLMResult` in function `core.agent.fc_agent_runner.FunctionCallAgentRunner.extract_blocking_tool_calls` [bad-argument-type]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:242:52
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:248:52
 ERROR Argument `None` is not assignable to parameter `query` with type `str` in function `core.agent.fc_agent_runner.FunctionCallAgentRunner._organize_user_query` [bad-argument-type]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:278:46
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:284:46
 ERROR Object of class `FunctionCallAgentRunner` has no attribute `session` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:353:58
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:311:58
 ERROR Object of class `FunctionCallAgentRunner` has no attribute `session` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:401:19
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:359:19
 ERROR Object of class `FunctionType` has no attribute `return_value` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:434:9
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:407:9
+ERROR Object of class `FunctionType` has no attribute `side_effect` [missing-attribute]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:444:9
+ERROR Object of class `FunctionCallAgentRunner` has no attribute `session` [missing-attribute]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:451:25
+ERROR Object of class `PromptMessage` has no attribute `opaque_body` [missing-attribute]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:461:16
+ERROR Object of class `FunctionType` has no attribute `assert_not_called` [missing-attribute]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:465:9
+ERROR `Literal['context_size']` is not assignable to dict key type `ModelPropertyKey` [bad-assignment]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:468:62
+ERROR `list[SimpleNamespace]` is not assignable to attribute `parameter_rules` with type `list[ParameterRule]` [bad-assignment]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:469:60
+ERROR Object of class `FunctionType` has no attribute `side_effect` [missing-attribute]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:472:9
+ERROR Object of class `FunctionType` has no attribute `return_value` [missing-attribute]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:484:9
+ERROR Object of class `FunctionCallAgentRunner` has no attribute `session` [missing-attribute]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:489:29
+ERROR Object of class `FunctionType` has no attribute `call_count` [missing-attribute]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:491:16
+ERROR Object of class `PromptMessage` has no attribute `opaque_body` [missing-attribute]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:492:16
+ERROR Object of class `FunctionType` has no attribute `call_args` [missing-attribute]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:494:16
+ERROR `Literal['context_size']` is not assignable to dict key type `ModelPropertyKey` [bad-assignment]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:509:62
+ERROR `list[SimpleNamespace]` is not assignable to attribute `parameter_rules` with type `list[ParameterRule]` [bad-assignment]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:510:60
+ERROR Object of class `FunctionType` has no attribute `return_value` [missing-attribute]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:514:9
+ERROR Object of class `FunctionType` has no attribute `return_value` [missing-attribute]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:522:13
+ERROR Object of class `AIModel` has no attribute `get_num_tokens` [missing-attribute]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:536:9
+ERROR Object of class `FunctionType` has no attribute `side_effect` [missing-attribute]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:561:9
+ERROR Object of class `FunctionCallAgentRunner` has no attribute `session` [missing-attribute]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:562:25
+ERROR Object of class `FunctionType` has no attribute `return_value` [missing-attribute]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:581:9
+ERROR Object of class `FunctionCallAgentRunner` has no attribute `session` [missing-attribute]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:583:29
+ERROR Object of class `FunctionType` has no attribute `call_count` [missing-attribute]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:585:16
+ERROR Object of class `FunctionType` has no attribute `return_value` [missing-attribute]
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:592:9
 ERROR Object of class `FunctionCallAgentRunner` has no attribute `session` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:436:35
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:594:35
 ERROR Object of class `FunctionType` has no attribute `call_args` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:438:33
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:596:33
 ERROR Object of class `FunctionType` has no attribute `call_args` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:439:33
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:597:33
 ERROR Object of class `FunctionType` has no attribute `call_args` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:440:16
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:598:16
 ERROR Object of class `FunctionType` has no attribute `assert_called` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:445:9
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:603:9
 ERROR Object of class `FunctionType` has no attribute `call_args_list` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:447:23
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:605:23
 ERROR Object of class `FunctionCallAgentRunner` has no attribute `session` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:453:19
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:611:19
 ERROR Argument `list[TextPromptMessageContent]` is not assignable to parameter `content` with type `str | None` in function `DummyMessage.__init__` [bad-argument-type]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:470:57
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:628:57
 ERROR Object of class `FunctionType` has no attribute `return_value` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:476:9
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:634:9
 ERROR Argument `list[TextPromptMessageContent]` is not assignable to parameter `content` with type `str | None` in function `DummyMessage.__init__` [bad-argument-type]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:492:57
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:650:57
 ERROR Object of class `FunctionType` has no attribute `side_effect` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:500:9
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:658:9
 ERROR Object of class `FunctionCallAgentRunner` has no attribute `session` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:502:35
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:660:35
 ERROR Argument `list[TextPromptMessageContent]` is not assignable to parameter `content` with type `str | None` in function `DummyMessage.__init__` [bad-argument-type]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:508:46
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:666:46
 ERROR Object of class `FunctionType` has no attribute `return_value` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:511:9
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:669:9
 ERROR Object of class `FunctionCallAgentRunner` has no attribute `session` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:513:35
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:671:35
 ERROR Object of class `FunctionType` has no attribute `call_args` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:515:16
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:673:16
 ERROR Object of class `FunctionType` has no attribute `return_value` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:531:9
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:689:9
 ERROR Object of class `FunctionCallAgentRunner` has no attribute `session` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:542:35
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:700:35
 ERROR Object of class `FunctionType` has no attribute `side_effect` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:558:9
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:716:9
 ERROR Object of class `FunctionCallAgentRunner` has no attribute `session` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:560:35
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:718:35
 ERROR Object of class `FunctionType` has no attribute `side_effect` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:575:9
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:733:9
 ERROR Object of class `FunctionType` has no attribute `return_value` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:580:9
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:738:9
 ERROR Object of class `FunctionCallAgentRunner` has no attribute `session` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:589:35
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:747:35
 ERROR Object of class `FunctionType` has no attribute `call_args_list` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:595:25
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:753:25
 ERROR Object of class `NoneType` has no attribute `max_iteration` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:599:9
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:757:9
 ERROR Object of class `FunctionType` has no attribute `return_value` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:611:9
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:769:9
 ERROR Object of class `FunctionCallAgentRunner` has no attribute `session` [missing-attribute]
-   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:614:29
+   --> tests/unit_tests/core/agent/test_fc_agent_runner.py:772:29
 ERROR Missing argument `description` in function `core.agent.plugin_entities.AgentStrategyEntity.__init__` [missing-argument]
    --> tests/unit_tests/core/agent/test_plugin_entities.py:281:32
 ERROR Invalid key for TypedDict `AppParametersDict`, got `Unknown` [bad-typed-dict-key]

github-actions · 2026-08-22T11:25:23Z

Pyrefly Type Coverage

Metric	Base	PR	Delta
Type coverage	67.36%	67.37%	+0.00%
Strict coverage	66.98%	66.99%	+0.00%
Typed symbols	51,259	51,272	+13
Untyped symbols	24,975	24,978	+3
Modules	3560	3561	+1

codecov · 2026-09-23T10:13:59Z

Codecov Report

❌ Patch coverage is 96.42857% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 88.06%. Comparing base (d1391c2) to head (29affd6).
⚠️ Report is 45 commits behind head on main.

Files with missing lines	Patch %	Lines
api/core/agent/fc_agent_runner.py	96.42%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #41095      +/-   ##
==========================================
+ Coverage   87.96%   88.06%   +0.10%     
==========================================
  Files        5267     5128     -139     
  Lines      324459   321848    -2611     
  Branches    65729    64690    -1039     
==========================================
- Hits       285398   283428    -1970     
+ Misses      33695    33048     -647     
- Partials     5366     5372       +6

Flag	Coverage Δ
api	`87.88% <96.42%> (+0.21%)`	⬆️
cli	`?`

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

crazywoola · 2026-09-28T02:13:32Z

Thank you for your contribution and the effort you’ve put into this PR. It currently has merge conflicts with main, so we’re closing it as part of our PR queue cleanup. If you’d like to continue, please update your branch against the latest main, resolve the conflicts, and reopen this PR or submit a new one. We appreciate your help improving Dify and would be happy to revisit the updated contribution.

hpiclaranet added 2 commits August 22, 2026 12:04

hpiclaranet mentioned this pull request Aug 22, 2026

fix(anthropic): round-trip thinking blocks through opaque_body (issue #3658) langgenius/dify-official-plugins#3715

Closed

fix(agent): preserve provider replay state across tool iterations

290aa59

WH-2099 self-assigned this Sep 23, 2026

This was referenced Sep 23, 2026

feat(openai): add GPT-6 models langgenius/dify-official-plugins#3918

Merged

feat(anthropic): add Claude Opus 5.5 langgenius/dify-official-plugins#3919

Merged

fix(agent): complete replay transport and preserve context budgets

f23fef0

github-actions Bot added web This relates to changes on the web. e2e End-to-end tests and E2E test infrastructure. labels Sep 23, 2026

WH-2099 changed the title ~~fix(agent): carry provider opaque_body state across function-calling agent iterations~~ fix(agent): preserve provider state and signed tool-loop prefixes Sep 23, 2026

WH-2099 mentioned this pull request Sep 23, 2026

feat: preserve opaque_body provider state on prompt messages and LLM results langgenius/graphon#261

Merged

2 of 3 tasks

test(plugin): type the provider replay transport regression

a637fc4

github-actions Bot added size:L This PR changes 100-499 lines, ignoring generated files. and removed size:L This PR changes 100-499 lines, ignoring generated files. web This relates to changes on the web. e2e End-to-end tests and E2E test infrastructure. labels Sep 23, 2026

fix(agent): retain small output budgets and complete replay dependency

29affd6

crazywoola closed this Sep 28, 2026

WH-2099 mentioned this pull request Oct 13, 2026

fix(llm): preserve opaque snapshots and content blocks in blocking invocation langgenius/dify-plugin-sdks#389

Closed

WH-2099 mentioned this pull request Oct 21, 2026

fix(agent): preserve complete model messages through tools and history #44502

Draft

3 of 4 tasks

Navigation Menu

Repository navigation

fix(agent): preserve provider state and signed tool-loop prefixes - #41095

fix(agent): preserve provider state and signed tool-loop prefixes#41095
hpiclaranet wants to merge 6 commits into
langgenius:mainfrom
hpiclaranet:fix/agent-opaque-body-round-trip

hpiclaranet commented Aug 22, 2026 •

edited by WH-2099

Loading

Uh oh!

hpiclaranet commented Aug 22, 2026

Uh oh!

github-actions Bot commented Aug 22, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Aug 22, 2026 •

edited

Loading

Uh oh!

codecov Bot commented Sep 23, 2026 •

edited

Loading

Uh oh!

crazywoola commented Sep 28, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

hpiclaranet commented Aug 22, 2026 • edited by WH-2099 Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Dependencies and release gate

Validation

Screenshots

Checklist

Uh oh!

hpiclaranet commented Aug 22, 2026

Uh oh!

github-actions Bot commented Aug 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Pyrefly Diff

Uh oh!

github-actions Bot commented Aug 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Pyrefly Type Coverage

Uh oh!

codecov Bot commented Sep 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

crazywoola commented Sep 28, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

hpiclaranet commented Aug 22, 2026 •

edited by WH-2099

Loading

github-actions Bot commented Aug 22, 2026 •

edited

Loading

github-actions Bot commented Aug 22, 2026 •

edited

Loading

codecov Bot commented Sep 23, 2026 •

edited

Loading