ichliebewxy/agent_console

How ichliebewxy/agent_console works

Agent Console is a local multi-agent and RAG workspace. The main workflow streams a user chat through the FastAPI chat route to the agent, which loads session history and may answer directly, use tools, search knowledge, or delegate to a skill agent; results and traces return over SSE and are persisted. Separate product paths manage document ingestion and hybrid retrieval, durable plan-and-execute runs, and long-term memory. The README documents capabilities beyond the sampled excerpts; those branches are included at supported module paths, while uncertain internal wiring is omitted.

Main entry points: agent.py Main Agent · routes_documents.py Document API · rag_pipeline.py RAG Pipeline · retrieval_steps.py Hybrid Retrieval

Components 26

Console and API

Web Console frontend/js/chat.js
FastAPI App backend/app.py
Chat Route backend/routes_chat.py

Agent and Tools

Main Agent backend/agent.py
Core Tools backend/core_tools.py
Reviewed Bash backend/bash_tool.py
Skill Delegation backend/subagents.py
Skills Catalog backend/skill_service.py
MCP Tools backend/mcp_service.py

Knowledge Base

Document API backend/routes_documents.py
Document Parsing backend/document_loader.py
Chunk Embeddings backend/embedding.py
Vector Index backend/milvus_client.py
Parent Chunks backend/parent_chunk_store.py
Knowledge Search Tool backend/search_tool.py
RAG Pipeline backend/rag_pipeline.py
Hybrid Retrieval backend/retrieval_steps.py
Expanded Retrieval backend/rag_expanded.py

Durable Workflows

Plan and Execute backend/workflow_graph.py
Run Checkpoints backend/checkpoint_service.py
Run Controls backend/routes_runs.py

Memory and State

Session History backend/conversation_storage.py
Long-term Memory backend/memory_service.py
Memory Extraction backend/memory_extraction.py
Session Artifacts backend/artifact_service.py

Other components

User

Connections 28

User → Web Console: sends message
Web Console → Chat Route: posts chat
Chat Route → Main Agent: invokes
Main Agent → Session History: loads and saves
Main Agent → Long-term Memory: searches context
Main Agent → Core Tools: dispatches tools
Core Tools → Reviewed Bash: uses guarded shell
Main Agent → Skill Delegation: delegates
Skill Delegation → Skills Catalog: loads skills
Main Agent → MCP Tools: invokes discovered tools
Main Agent → Plan and Execute: starts complex runs
Plan and Execute → Run Checkpoints: checkpoints state
Run Controls → Run Checkpoints: inspects and resumes
Run Controls → Session History: projects run state
Main Agent → Knowledge Search Tool: calls search tool
Knowledge Search Tool → RAG Pipeline: runs query
RAG Pipeline → Hybrid Retrieval: retrieves documents
RAG Pipeline → Expanded Retrieval: retries expanded query
Hybrid Retrieval → Vector Index: hybrid searches
Hybrid Retrieval → Parent Chunks: loads parent context
Document API → Document Parsing: parses uploads
Document API → Chunk Embeddings: chunks and embeds
Document API → Vector Index: writes leaf vectors
Document API → Parent Chunks: writes parent chunks
Main Agent → Memory Extraction: starts after reply
Memory Extraction → Long-term Memory: deduplicates and writes
Main Agent → Session Artifacts: lists session outputs
Main Agent → Web Console: streams response and trace

More Python diagrams: heidelberg-hepml/lorentz-gatr · henry0hai/browser-n8n-local · heqiang-huang/yolo-ts · ilearn-lab/evoharness · insertx2k/temp_cleaner_gui

Drawn by GitDiagram on September 29, 2026 from the repository's file tree, README and source files. Read as Markdown ·