How denis2054/context-engineering-for-multi-agent-systems works
This repository is a collection of Context Engine implementations and deployment editions, rather than one single runtime. The principal workflow accepts a goal, plans agent work, executes retrieval and generation, and returns an output with an execution trace. The graph preserves the classic engine, LangChain edition, NIM DAG editions, and documented observability and computer-use capabilities. Some subsystem links rely on README statements or SOURCE INDEX references; unsampled behavior is not expanded beyond that evidence.
Main entry points: lc_engine.py LangChain planner and executor · engine_nim.py NIM DAG orchestrator · main.py Deployment API · engine.py Classic orchestrator
Components 21
Core Context Engine
- Classic orchestrator
commons/engine.py - Retrieval and writer agents
commons/agents.py - Agent registry
commons/registry.py - LLM and retrieval helpers
commons/helpers.py - Pinecone index
LangChain Edition
- LangChain planner and executor
langchain/lc_engine.py - LangChain engine helpers
langchain/lc_helpers.py - LangChain agents
langchain/lc_agents.py
NIM DAG Editions
- NIM DAG orchestrator
nim/engine_nim.py - NIM agents
nim/agents_nim.py - NIM agent registry
nim/registry_nim.py - NIM DAG executor
nim/run_dag_nim.py - Deployment API
nemotron_docker_railway/main.py - Deployment orchestrator
nemotron_docker_railway/engine.py - Gate harness
nemotron_docker_railway/harness.py - Pinecone adapter
nemotron_docker_railway/adapters.py - DAG executor
nemotron_docker_railway/run_dag.py
Trace and UI
- LangChain trace dashboard
langchain/lc_dashboard.py - NIM trace dashboard
nim/dashboard_nim.py
Specialized Capabilities
- Astra computer use
gpt-6-astra/astra.py
Other components
- Goal requester
Connections 22
- Goal requester → Classic orchestrator: submits goal
commons/engine.py - Classic orchestrator → LLM and retrieval helpers: plans and retrieves
commons/engine.py - Classic orchestrator → Agent registry: resolves agents
commons/engine.py - Agent registry → Retrieval and writer agents: registers handlers
commons/registry.py - Retrieval and writer agents → LLM and retrieval helpers: retrieves and generates
commons/agents.py - LLM and retrieval helpers → Pinecone index: queries vectors
commons/helpers.py - Goal requester → LangChain planner and executor: submits goal
README.md - LangChain planner and executor → LangChain engine helpers: uses helpers
langchain/lc_engine.py - LangChain planner and executor → LangChain agents: dispatches agents
- LangChain planner and executor → LangChain trace dashboard: emits trace for
README.md - Goal requester → NIM DAG orchestrator: submits goal
README.md - NIM DAG orchestrator → NIM DAG executor: executes DAG through
nim/engine_nim.py - NIM DAG orchestrator → NIM agents: dispatches agents
- NIM agent registry → NIM agents: registers handlers
nim/registry_nim.py - NIM DAG orchestrator → NIM trace dashboard: provides trace for
README.md - Goal requester → Deployment API: posts goal
nemotron_docker_railway/main.py - Deployment API → Deployment orchestrator: invokes engine
nemotron_docker_railway/main.py - Deployment API → Gate harness: configures gates
nemotron_docker_railway/main.py - Deployment API → Pinecone adapter: configures storage
nemotron_docker_railway/main.py - Deployment orchestrator → DAG executor: executes DAG through
nemotron_docker_railway/engine.py - Deployment orchestrator → Pinecone adapter: uses storage adapter
nemotron_docker_railway/engine.py - Goal requester → Astra computer use: requests computer use
README.md