How pashaeimehran/agentic-ai-job-matcher works
This is an agentic job-matching product: job seekers search German postings, retrieve CV evidence, receive deterministic requirement assessments and scores, and can save jobs or ask an agent to explain results. REST, MCP, the LangGraph agent, and a manual baseline expose shared tools. The graph preserves extraction, hybrid CV retrieval, matching, persistence, and LLM operations. Edges reflect README statements and shown source references/calls; several tool registrations and subsystem internals were not sampled, so those connections rely on the README rather than claims of inspected implementation.
Main entry points: registry.py Tool registry · service.py CV retrieval service · app.py FastAPI · instrumented.py LLM instrumentation
Components 24
Interfaces and agents
- FastAPI
src/api/app.py - MCP server
src/mcp_server/server.py - LangGraph agent
src/graph/agent.py - Graph nodes
src/graph/nodes.py - Manual agent loop
src/agent/loop.py
Job discovery
- Job catalog
src/tools/jobs.py - BA job source
src/data_sources/ba_jobsuche.py - Job extraction
src/extraction/extractor.py
CV and matching
- CV retrieval service
src/rag/service.py - CV ingestion
src/rag/ingestion.py - Embeddings
src/rag/embeddings.py - Hybrid retrieval
src/matching/retrieval.py - Evidence matching
src/matching/service.py - Deterministic scoring
src/matching/scoring.py
Shared tools and state
- Tool registry
src/tools/registry.py - Tool context
src/tools/context.py - Saved jobs
src/tools/saved_jobs.py - PostgreSQL / pgvector
src/rag/postgres.py
LLM and operations
- LLM provider
src/llm/provider.py - LLM instrumentation
src/observability/instrumented.py - Tracing
src/observability/tracing.py
Other components
- Job seeker
- REST client
- MCP client
Connections 27
- Job seeker → FastAPI: uses REST
README.md - REST client → FastAPI: calls endpoints
README.md - MCP client → MCP server: calls tools
README.md - FastAPI → LangGraph agent: routes chat
README.md - FastAPI → Tool registry: calls tools
README.md - MCP server → Tool registry: calls tools
README.md - LangGraph agent → Graph nodes: builds graph
src/graph/agent.py - Graph nodes → Tool registry: executes tools
src/graph/nodes.py - Manual agent loop → Tool registry: executes tools
src/agent/loop.py - Tool registry → Job catalog: dispatches search
README.md - Tool registry → CV retrieval service: dispatches retrieval
README.md - Tool registry → Evidence matching: dispatches matching
README.md - Tool registry → Saved jobs: dispatches save
README.md - BA job source → Job extraction: provides postings
- CV retrieval service → CV ingestion: ingests and chunks
src/rag/service.py - CV retrieval service → Embeddings: embeds text
src/rag/service.py - CV retrieval service → PostgreSQL / pgvector: stores and searches
- Evidence matching → Hybrid retrieval: retrieves evidence
src/matching/service.py - Hybrid retrieval → CV retrieval service: searches CV
src/matching/retrieval.py - Evidence matching → Deterministic scoring: scores requirements
src/matching/service.py - Tool context → CV retrieval service: provides service
src/tools/context.py - Tool context → Job catalog: provides catalog
src/tools/context.py - Saved jobs → PostgreSQL / pgvector: persists jobs
README.md - LangGraph agent → LLM instrumentation: uses provider
README.md - Manual agent loop → LLM instrumentation: uses provider
README.md - LLM instrumentation → LLM provider: wraps provider
src/observability/instrumented.py - LLM instrumentation → Tracing: records calls
src/observability/instrumented.py