fix: add mcp components #1141
Draft
fix: add mcp components #1141
Conversation
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
…age and mcp to be discussed; the extraction logic not done yet; the config load is ugly) init commit 2
- Refine context7.py prompt - Handle/pause on timeout - Optimize context7 logic - Pass content downstream when enabled - Add PyTorch Image Models (timm) search to index.ts
Hoder-zyf
changed the title
you-n-g
reviewed
Aug 4, 2025
- Implement litellm integration for MCP extra configuration - Enable flexible model calling through litellm adapter Next steps: - [ ] Multi-MCP registration and validation - [ ] Eval module testing - [ ] Code refactoring for better readability - [ ] README documentation
Replace dual-function API (query_mcp/query_mcp_auto) with single flexible interface supporting three modes: auto (all services), single service, and multi-service parallel processing. Key improvements: - Remove API complexity: one function instead of two - Enable true parallelism: services run concurrently, not sequentially - Add MultiServiceContext for efficient connection management - Fix LiteLLM tool calling edge cases with proper finish_reason handling - Update examples demonstrating new unified interface Performance impact: Multi-service queries now execute in parallel rather than trying services one by one, significantly reducing response time. Breaking change: query_mcp_auto() removed, use query_mcp() instead.
- Add verbose parameter to control logging output levels
- Default (verbose=False): Show only essential logs
- Verbose mode (verbose=True): Show detailed debugging info
- Enhance error detection with comprehensive pattern matching
- Add rate limit error detection ('rate limit', 'too many
requests')
- Add content availability errors ('no content available')
- Fix MCP services returning errors as successful responses
- Refactor error handling with data-driven approach
- Consolidate error patterns into ERROR_DEFINITIONS constant
- Eliminate code duplication in error detection functions
- Add priority-based error message formatting
- Remove cost calculation logging from LiteLLM backend
- Clean up unused imports and improve code organization
you-n-g
reviewed
Sep 1, 2025
you-n-g
reviewed
Sep 1, 2025
mcp_config.json.example
Outdated
| "mcpServices": { | ||
| "context7": { | ||
| "url": "http://localhost:8123/mcp", | ||
| "handler": "rdagent.components.mcp.context7.handler:Context7Handler", |
Contributor
There was a problem hiding this comment.
mix the API Handler and the prompt customization.
Contributor
There was a problem hiding this comment.
Two different responsibility.
- API Handler (1 - 1)
- prompt customization. ( Agent level custmization, Many-to-many )
Contributor
There was a problem hiding this comment.
MCPAgent + MCPService( API Handler is included)
improve error handling
Hoder-zyf
changed the title
Hoder-zyf
force-pushed
the
yifei/context7
branch
from
3e3cee1 to
8bfc09c
Compare
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.
Description
This PR integrates a new MCP (Model Context Protocol) Context7 capability and wires it into the coder/data-science pipeline so the agent can search relevant documentation when encountering implementation errors.
Key changes
rdagent.components.mcppackage with a dedicatedcontext7module, including client, prompts, examples, and README, enabling MCP-based doc search.rdagent/components/coder/data_science/pipeline/eval.pyand itsprompts.yaml.rdagent/app/data_science/conf.pyto make Context7 usage configurable; control via theCONTEXT7_ENABLEDenvironment variable. Also adjusts the hypothesis critique toggle default.rdagent/components/coder/CoSTEER/evaluators.pyto align with the pipeline changes..devcontainer/devcontainer.jsonand.gitignore.Why it matters
Configuration
CONTEXT7_ENABLED=1andMCP_ENABLED=TrueMotivation and Context
Agent runs often fail due to missing APIs, library nuances, or environment quirks. Integrating an MCP-driven doc search lets the system proactively fetch relevant references based on the current error, guiding the next fix step. The added config flag ensures teams can roll this out gradually and A/B the impact.
Types of changes