Feat/agentcore adapter#18
Open
rkmaws wants to merge 8 commits into
Open
Conversation
added 8 commits
April 24, 2026 23:15
Minimal changes to existing community code to support the AgentCore deployment interface. All changes are additive and backwards-compatible with the existing ECS/Docker deployment. - .gitignore: add AgentCore artifacts (.bedrock_agentcore/, Dockerfile, *.db, .unique-id-*) - pyproject.toml: register agentcore pytest marker, add test deps - product_setup_flow.py: add CSV adapter fallback for loading products when MCP server is not available (AgentCore runs FastAPI+MCP in-process) - chat/main.py: add storage parameter to ChatInterface constructor for external storage backend injection (AgentCore uses SQLite in-memory)
Add Amazon Bedrock AgentCore deployment interface for the seller agent. Wraps the existing PublisherCrew and ChatInterface without modifying community-maintained agent/crew code — all new files live in src/ad_seller/interfaces/agentcore/ and patches/. Runtime architecture: - HTTP mode: BedrockAgentCoreApp entrypoint with two routing paths - crew: CrewAI PublisherCrew with native Bedrock Converse LLM - chat: existing ChatInterface keyword router - MCP mode: FastMCP server exposing all 41 tools via Streamable HTTP - Background FastAPI+MCP server on localhost for CrewAI tool callbacks Key components: - http_main.py: AgentCore entrypoint, routing, crew setup - crew_tools.py: BaseTool subclasses for inventory, pricing, deals - mcp_main.py: MCP-only entrypoint for tool serving - patches/crewai_bedrock_fix.py: Bedrock Converse API compatibility (orphaned toolUse/toolResult sanitization, arg extraction fix) - deploy.sh: Build and deploy via agentcore CLI with CodeBuild - Workshop demo data: synthetic Meridian Media Group inventory (CSV) Tests: - 209 unit tests (tools, routing,community-maintained agent/crew code — all new files live in src/mesrc/ad_seller/interfaces/agentcore/ and patches/. Runtime arnt Runtime architecture: - HTTP mode: BedrockAgentnit- HTTP mode: Bedrockcr - crew: CrewAI PublisherCrew with native Bedrock Converse LLM - chat: existing ChatInterface keyword router - MCP mode: FastMCP server exposing all 41 tools via Streamable HTTP - Background FastAPI+MCP server on localhost for CrewAI tool callbacks Key components: - http_main.py: AgentCore entrypoint, routing, crew setup - crew_tools.py: BaseTool subclasses for inventory, pricing, deals - mcp_main.py: MCP-only entrypoint for tool serving - patches/crewai_bedrock_fix.py: Bedrock Converse API compatibility (orphaned toolUse/toolResult sanitization, arg extraction fix) - deploy.sh: Build and deploy via agentcore CLI with CodeBuild - Workshop demo data: synthetic Meridian Media Group inventory (CSV) Tests: - 209 unit tests (tools, routing, patches, deploy artifacts, data) - 9 integration tests (live runtime invocation via agentcore CLI) All changes in interfaces/agentcore/ and patches/ — no modifications to community-maintained agent, crew, or flow code.
- Memory patch: replace StorageBackend with AgentCoreStorageBackend - Set _read_only=True to prevent RememberTool injection - Apply patch in http_main.py when CREW_MEMORY_ENABLED=true - Memory LLM: Nova Lite, Crew LLM: Nova Pro (unchanged) - Deploy script: CREW_MEMORY_ENABLED=true + MEMORY_LLM_MODEL
# Conflicts: # .gitignore # pyproject.toml
- crew_tools.py: CreateDealTool fallback now initializes chat interface to access CSV-loaded products (fixes 'Product not found' when REST static catalog doesn't have CSV products) - mcp_server.py: Accept int|None on limit/days params to handle Bedrock Converse sending null for optional tool arguments - docs/PRODUCTION_DATA_INTEGRATION.md: Guide for replacing demo CSV/SQLite with production ad server + database backends
- New: s3_csv_adapter.py — reads inventory/audiences from S3 with glob merge + 5-min cache - New: storage-s3.yaml — CloudFormation template for seller data bucket + IAM - New: tests/unit/clients/test_s3_csv_adapter.py — 9 unit tests (all passing) - deploy.sh: added --inventory csv|s3|gam|freewheel flag (separated from --storage) - deploy.sh: provision_s3_data_bucket() creates bucket + uploads CSVs - ad_server_base.py: added S3 enum + factory case - settings.py: added s3_data_bucket, s3_data_prefix, s3_data_region - http_main.py: _get_chat() now uses get_ad_server_client() (respects AD_SERVER_TYPE) - s3_csv_adapter.py: attaches raw dict (floor_price_cpm, inventory_type) to items Note: ProductSetupFlow in the FastAPI background server still needs S3 wiring (follow-up)
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
1 participant
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.
Amazon Bedrock Agentcore integration