Fix CI server pull_request failure in Mistral component snapshot test by ivicac · Pull Request #4939 · bytechefhq/bytechef

ivicac · 2026-05-08T05:24:01Z

Description

Resolved the failing GitHub Actions job CI Build on PULL Request / server (pull_request) by fixing a stale test snapshot in the Mistral component module.

The failure was caused by MistralComponentHandlerTest expecting a top-level "inputs": null field that is no longer present in the serialized component definition.
Updated server/libs/modules/components/ai/llm/mistral/src/test/resources/definition/mistral_v1.json to match current output.

Type of change

Bug fix (non-breaking change which fixes an issue)

How Has This Been Tested?

Ran ./gradlew :server:libs:modules:components:ai:llm:mistral:test --no-daemon
Verified MistralComponentHandlerTest passes and build is successful

Checklist:

My code follows the style guidelines of this project
I have performed a self-review of my own code
I have commented my code, particularly in hard-to-understand areas
I have made corresponding changes to the documentation
My changes generate no new warnings
I have added tests that prove my fix is effective or that my feature works
New and existing unit tests pass locally with my changes

Copilot

Pull request overview

This PR updates the AI Agent component to wire Spring AI tool-calling support via ToolCallingManager/ToolCallAdvisor, and introduces explicit advisor ordering so chat-memory advisors run with a defined precedence.

Changes:

Inject ToolCallingManager through the AI Agent component handler/actions and add ToolCallAdvisor to the advisor chain.
Define explicit advisor ordering for chat-memory advisors (Prompt/VectorStore variants).
Update unit tests to match new constructors/signatures.

Reviewed changes

Copilot reviewed 15 out of 15 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
server/libs/modules/components/ai/agent/src/test/java/com/bytechef/component/chat/AiAgentComponentHandlerTest.java	Updates handler construction for the new constructor signature.
server/libs/modules/components/ai/agent/src/test/java/com/bytechef/component/ai/agent/action/AbstractAiAgentChatActionTest.java	Adds `ToolCallingManager` mocking and updates action construction.
server/libs/modules/components/ai/agent/src/main/java/com/bytechef/component/ai/agent/AiAgentComponentHandler.java	Injects `ToolCallingManager` and wires it into chat/stream chat actions.
server/libs/modules/components/ai/agent/src/main/java/com/bytechef/component/ai/agent/action/AiAgentStreamChatAction.java	Threads `ToolCallingManager` into the stream chat action.
server/libs/modules/components/ai/agent/src/main/java/com/bytechef/component/ai/agent/action/AiAgentChatAction.java	Threads `ToolCallingManager` into the chat action and updates factory method signature.
server/libs/modules/components/ai/agent/src/main/java/com/bytechef/component/ai/agent/action/AbstractAiAgentChatAction.java	Adds `ToolCallAdvisor` wiring and adjusts advisor composition around chat memory.
server/libs/modules/components/ai/agent/chat-memory/chat-memory-vectorstore/src/main/java/com/bytechef/component/ai/chat/memory/cluster/VectorStoreChatMemory.java	Sets explicit advisor order for vectorstore chat memory advisor.
server/libs/modules/components/ai/agent/chat-memory/chat-memory-redis/src/main/java/com/bytechef/component/ai/agent/chat/memory/redis/cluster/RedisChatMemory.java	Sets explicit advisor order for redis chat memory advisor.
server/libs/modules/components/ai/agent/chat-memory/chat-memory-neo4j/src/main/java/com/bytechef/component/ai/agent/chat/memory/neo4j/cluster/Neo4jChatMemory.java	Sets explicit advisor order for neo4j chat memory advisor.
server/libs/modules/components/ai/agent/chat-memory/chat-memory-mongodb/src/main/java/com/bytechef/component/ai/agent/chat/memory/mongodb/cluster/MongoDbChatMemory.java	Sets explicit advisor order for mongodb chat memory advisor.
server/libs/modules/components/ai/agent/chat-memory/chat-memory-jdbc/src/main/java/com/bytechef/component/ai/agent/chat/memory/jdbc/cluster/JdbcChatMemory.java	Sets explicit advisor order for jdbc chat memory advisor.
server/libs/modules/components/ai/agent/chat-memory/chat-memory-in-memory/src/main/java/com/bytechef/component/ai/agent/chat/memory/memory/cluster/InMemoryChatMemory.java	Sets explicit advisor order for in-memory chat memory advisor.
server/libs/modules/components/ai/agent/chat-memory/chat-memory-cosmosdb/src/main/java/com/bytechef/component/ai/agent/chat/memory/cosmosdb/cluster/CosmosDbChatMemory.java	Sets explicit advisor order for cosmosdb chat memory advisor.
server/libs/modules/components/ai/agent/chat-memory/chat-memory-cassandra/src/main/java/com/bytechef/component/ai/agent/chat/memory/cassandra/cluster/CassandraChatMemory.java	Sets explicit advisor order for cassandra chat memory advisor.
server/libs/modules/components/ai/agent/chat-memory/chat-memory-builtin/src/main/java/com/bytechef/component/ai/agent/chat/memory/builtin/cluster/ChatMemory.java	Sets explicit advisor order for builtin chat memory advisor.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

marko-kriskovic · 2026-05-19T14:55:43Z

need to wait for #5009 before merging

gitguardian · 2026-06-03T12:22:43Z

️✅ There are no secrets present in this pull request anymore.

If these secrets were true positive and are still valid, we highly recommend you to revoke them.
While these secrets were previously flagged, we no longer have a reference to the
specific commits where they were detected. Once a secret has been leaked into a git
repository, you should consider it compromised, even if it was deleted immediately.
Find here more information about risks.

^{_{🦉 GitGuardian detects secrets in your source code to help developers and security teams secure the modern development process. You are seeing this because you or someone else with access to this repository has authorized GitGuardian to scan your pull request.}}

…visor wiring Standardize (AiAgentToolFacade, ClusterElementDefinitionService, ToolCallingManager) across AiAgentChatAction.of, AiAgentStreamChatAction.of, their constructors, and the AiAgentComponentHandler call sites. Add tests asserting ToolCallAdvisor is present in the advisor list and that disableInternalConversationHistory() is applied only when a chat-memory advisor is configured. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…participates in every tool-call iteration The previous setup ran every chat-memory advisor at HIGHEST_PRECEDENCE + 200 and ToolCallAdvisor at HIGHEST_PRECEDENCE + 300 — Spring's lower-order-runs-first contract meant memory ran ONCE outside the tool-call loop. With disableInternalConversationHistory() enabled, ToolCallAdvisor passes only [system, lastMessage] between loop iterations and expects a downstream ChatMemoryAdvisor to supply the preceding assistant-with-tool-calls. Because no advisor was downstream, iteration 2 reached the model as just [system, toolResultMsg], producing the OpenAI 400 "messages with role 'tool' must be a response to a preceding message with 'tool_calls'" error. Drop the explicit orders on both sides so each advisor falls back to its Spring AI default: ToolCallAdvisor at HIGHEST_PRECEDENCE + 300 (its DEFAULT_ORDER) and the seven non-vector chat-memory advisors at DEFAULT_CHAT_MEMORY_PRECEDENCE_ORDER (Integer.MIN_VALUE + 1000). The vector-store advisor already uses the default after commit 37fb7aa8ae3. Spring AI documents these defaults as designed to compose — ToolCallAdvisor.DEFAULT_ORDER's Javadoc states "Placed early in the chain so that all downstream advisors (e.g. BaseChatMemoryAdvisor) participate in every tool-call iteration." Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

… to ToolCallAdvisor testChatMemoryAdvisorOrderedDownstreamOfToolCallAdvisor builds the agent's advisor chain end-to-end with a production-style MessageChatMemoryAdvisor (built via the Spring AI builder default, the same way every chat-memory component now constructs it) and asserts its order is greater than the ToolCallAdvisor's. Verified by hand that this test fails when the previous buggy ordering is restored — catching the exact regression that produced OpenAI's "messages with role 'tool' must be a response to a preceding message with 'tool_calls'" failure. testSpringAiDefaultOrdersComposeCorrectly pins the Spring AI defaults directly (ToolCallAdvisor.builder().build().getOrder() vs MessageChatMemoryAdvisor.builder(...).build().getOrder()) so a future spring-ai version bump that inverts those defaults fails the test at the bump PR instead of producing chat-memory breakage in production. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

… manually-registered ToolCallAdvisor Spring AI 2.0's docs make this call look optional or removable — auto-registered ToolCallAdvisors get the auto-disable behavior when a downstream ChatMemoryAdvisor is detected. Manually-registered ones (which we need because SuspendableToolCallingManager wraps the default ToolCallingManager) skip that path entirely, so the explicit call is load-bearing. Comment the WHY so a future reader doesn't reach for the obvious-looking deletion. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

M7 removes spring-ai-model-chat-memory-repository-cosmos-db (#6080), so delete the chat-memory-cosmosdb component along with its docs row and generated reference page. M7's "Sanitize Spring Boot related dependencies" (#6088) also stops leaking spring-boot-jdbc and spring-boot-autoconfigure transitively, so declare them explicitly in the two modules that were relying on the leak. The custom ToolCallAdvisor registration in AbstractAiAgentChatAction stays as-is: manual registration suppresses M7's new auto-registration (#5459), and SuspendableToolCallingManager requires the explicit toolCallingManager wiring that auto-registration cannot provide. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…AI M7 ToolSpec API Spring AI 2.0.0-M7 deprecated and marked for removal both ChatClient.ChatClientRequestSpec.toolCallbacks(List) and ChatClient.ChatClientRequestSpec.toolContext(Map) in favor of a unified tools(Consumer<ToolSpec>) entry point (see PR #6085 "Introduce ToolSpec fluent API"). This branch has a single deprecated call site — the toolCallbacks() registration in AbstractAiAgentChatAction — which is replaced with the new tools(spec -> spec.callbacks(...)) form. Behavior is identical; the new API just consolidates tool registration into a single fluent surface. (The toolContext() migrations from the upstream change live in suspend/resume code that is not part of this branch.) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

….0.0-M3 Two related fixes for the recently bumped Spring AI 2.0.0-M7: 1) Spring AI M7 packaging oversight. The M7 spring-ai-starter-mcp-client POM no longer depends on spring-ai-autoconfigure-mcp-client-common, but spring-ai-autoconfigure-mcp-client-httpclient:2.0.0-M7 still references classes in org.springframework.ai.mcp.client.common. autoconfigure.* (McpSseClientProperties, NamedClientMcpTransport, etc.). Boot fails at the bean-definition phase with ClassNotFoundException for McpSseClientProperties. The M7 jar exists on Maven Central (just not listed in the directory index) so an explicit dependency on spring-ai-autoconfigure-mcp-client-common pulls it back in. Added with a comment explaining the workaround so it can be removed once a future Spring AI release re-includes the transitive dependency. 2) MCP SDK 2.0.0-M3 sync. The version pin in libs.versions.toml was still M2 even though Spring AI M7 already resolves mcp-core to M3 transitively. Bumped to M3 explicitly to remove the mixed-version state, and updated the forked FilterableMcpAsyncServer.java to match the upstream M3 changes: - DefaultMcpStreamableServerSessionFactory constructor now takes a trailing jsonSchemaValidator parameter - addTool() validates the tool's inputSchema / outputSchema via jsonSchemaValidator.assertConforms() before insertion - McpSchema.TextContent / ListToolsResult / ListResourcesResult / ListResourceTemplatesResult / ListPromptsResult constructors are replaced with their builder().build() forms (the constructors are no longer public in M3) - Javadoc header bumped from "fork of McpAsyncServer (2.0.0-M2)" to "(2.0.0-M3)" Verified with bootRun: ServerApplication now starts cleanly in ~19s. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

M8 is a small stabilization release (4 bug fixes, 4 minor features); no API breakage, no module coordinate changes. Vendored spring-ai-tool-search-tool fork's super(...) patch remains valid since ToolCallAdvisor signature is unchanged in M8. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

@Serviceconnection

…pplicationIntTest worker-app has no JDBC dependency — it's stateless and reaches database state through *-remote-client modules. The vestigial @import predates the remote-client refactor. Under Spring Boot 4, @Serviceconnection now throws when no matching ConnectionDetails registrar is on the classpath, so importing a Postgres container in an app without JDBC autoconfig fails the context load. Removing the unused import restores green testIntegration without affecting sibling app tests (connection-app, execution-app, etc.) that legitimately pull in JDBC starters. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…AI M8 Snapshots drifted because the LLM components pull model lists dynamically from Spring AI's provider enums (GoogleGenAiChatModel.ChatModel.values(), MistralAiApi.ChatModel.values(), OpenAiApi.ChatModel.values()), and M8 refreshed those enums: - Gemini: dropped -preview suffix on flash-lite, added gemini-3.5-flash - OpenAI: gained ~3 new chat models in one action - Mistral: trimmed one retired model The bulk of the diff is whitespace from Jackson 3.x's default pretty-printer emitting " : " (with leading space) instead of ": "; net model changes are small. Other LLM components (anthropic, ollama, azure-open-ai, amazon-bedrock, stability) verified unaffected. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…AI M8 Snapshots drifted because the LLM components pull model lists dynamically from Spring AI's provider enums (GoogleGenAiChatModel.ChatModel.values(), MistralAiApi.ChatModel.values(), OpenAiApi.ChatModel.values()), and M8 refreshed those enums: - Gemini: dropped -preview suffix on flash-lite, added gemini-3.5-flash - OpenAI: gained ~3 new chat models in one action - Mistral: trimmed one retired model The bulk of the diff is whitespace from Jackson 3.x's default pretty-printer emitting " : " (with leading space) instead of ": "; net model changes are small. Other LLM components (anthropic, ollama, azure-open-ai, amazon-bedrock, stability) verified unaffected. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> mistral

…ConversationHistory()

…tform

This reverts commit 77f054e8a9bb29827e67673c71f80632d4906069.

…InternalConversationHistory()" This reverts commit 0be576f.

…nalConversationHistory

sonarqubecloud · 2026-06-06T09:55:26Z

Quality Gate passed for 'client'

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

ivicac requested a review from Copilot May 8, 2026 05:24

Copilot started reviewing on behalf of ivicac May 8, 2026 05:24 View session

Copilot AI reviewed May 8, 2026

View reviewed changes

ivicac requested a review from marko-kriskovic May 8, 2026 14:52

ivicac force-pushed the 1652 branch 2 times, most recently from edb55d1 to a400eee Compare May 17, 2026 09:48

marko-kriskovic force-pushed the 1652 branch 2 times, most recently from c6bd156 to a400eee Compare May 19, 2026 14:44

ivicac force-pushed the 1652 branch from 58efbe7 to 6aa815e Compare May 20, 2026 10:03

marko-kriskovic approved these changes May 22, 2026

View reviewed changes

ivicac force-pushed the 1652 branch from 8ef126a to 824e1cd Compare May 23, 2026 10:25

marko-kriskovic force-pushed the 1652 branch from 824e1cd to 6acc222 Compare May 25, 2026 11:05

ivicac force-pushed the 1652 branch 2 times, most recently from b7c086f to a6b8ee3 Compare June 3, 2026 12:22

This comment has been minimized.

Sign in to view

ivicac force-pushed the 1652 branch 2 times, most recently from e455f18 to cfedcf1 Compare June 3, 2026 13:27

Copilot started work on behalf of ivicac June 3, 2026 13:55 View session

Copilot AI changed the title ~~1652 Order advisors and wire ToolCallAdvisor in AI agent~~ Fix FTP integration test container flake in CI server build Jun 3, 2026

Copilot finished work on behalf of ivicac June 3, 2026 14:04

ivicac force-pushed the 1652 branch 2 times, most recently from 1bb2823 to 030c7ee Compare June 3, 2026 14:12

Copilot started work on behalf of ivicac June 3, 2026 14:31 View session

Copilot finished work on behalf of ivicac June 3, 2026 14:39

marko-kriskovic changed the title ~~Fix FTP integration test container flake in CI server build~~ Order advisors and wire ToolCallAdvisor in AI agent Jun 3, 2026

ivicac force-pushed the 1652 branch from 156574e to 2eda9f1 Compare June 4, 2026 16:53

Copilot started work on behalf of ivicac June 4, 2026 19:03 View session

Copilot AI changed the title ~~Order advisors and wire ToolCallAdvisor in AI agent~~ Fix CI server pull_request failure in Mistral component snapshot test Jun 4, 2026

marko-kriskovic force-pushed the 1652 branch 2 times, most recently from 9b62519 to c38a5ad Compare June 5, 2026 13:51

ivicac and others added 21 commits June 6, 2026 11:26

1652 Order advisors and wire ToolCallAdvisor in AI agent

9bbf531

1652 - ordering by DEFAULT_CHAT_MEMORY_PRECEDENCE_ORDER

770e7bf

1652 - workaround for deleting toolCallAdvisorBuilder.disableInternal…

18daea6

…ConversationHistory()

1652 - changed deprecated JsonParser

5e2c604

test: update mistral component snapshot for current definition

3e512e6

1570 Move workflow validate and instruction MCP tools to mcp-tool-pla…

6d24bb4

…tform

1652 - using filter in Chat memory directly

7f86f63

Revert "1652 - using filter in Chat memory directly"

245e5ac

This reverts commit 77f054e8a9bb29827e67673c71f80632d4906069.

Revert "1652 - workaround for deleting toolCallAdvisorBuilder.disable…

1ba97cb

…InternalConversationHistory()" This reverts commit 0be576f.

1652 - remembering internal tool conversation history on disableInter…

059d4e2

…nalConversationHistory

ivicac force-pushed the 1652 branch from c38a5ad to 059d4e2 Compare June 6, 2026 09:38

ivicac added 3 commits June 6, 2026 11:43

1652 - Fix test

2f3ee5b

4096 client - SF

7f15143

1652 SF

cbf9250

ivicac merged commit c6f7f32 into master Jun 6, 2026
6 of 7 checks passed

ivicac deleted the 1652 branch June 6, 2026 09:48

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix CI server pull_request failure in Mistral component snapshot test#4939

Fix CI server pull_request failure in Mistral component snapshot test#4939
ivicac merged 24 commits into
masterfrom
1652

ivicac commented May 8, 2026 •

edited by Copilot AI

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

marko-kriskovic commented May 19, 2026

Uh oh!

gitguardian Bot commented Jun 3, 2026 •

edited

Loading

Uh oh!

This comment has been minimized.

Uh oh!

sonarqubecloud Bot commented Jun 6, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

ivicac commented May 8, 2026 • edited by Copilot AI Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Type of change

How Has This Been Tested?

Checklist:

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

marko-kriskovic commented May 19, 2026

Uh oh!

gitguardian Bot commented Jun 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

️✅ There are no secrets present in this pull request anymore.

Uh oh!

This comment has been minimized.

Uh oh!

sonarqubecloud Bot commented Jun 6, 2026

Quality Gate passed for 'client'

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

ivicac commented May 8, 2026 •

edited by Copilot AI

Loading

gitguardian Bot commented Jun 3, 2026 •

edited

Loading