Developer AI Agent Duo
Built at Dassault Systèmes as part of the developer tooling initiative, this system pairs two specialized AI agents behind a single conversational interface. The Action Agent uses LangChain to execute real operations on the developer's behalf — triggering sandbox deployments, running CI/CD builds, and orchestrating multi-step pipeline commands through natural language, removing the overhead of memorizing CLI flags and navigating complex UIs. The Documentation Agent is a RAG pipeline that ingests internal development docs, chunks them with semantic-boundary-aware strategies, and embeds them for vector similarity search. Together, the duo lets engineers ask "deploy my branch to staging" and "how does the auth middleware work?" in the same conversation window.
Key Highlights
- 01Two-agent architecture — Action Agent for CI/CD operations, Documentation Agent for context-aware Q&A
- 02Action Agent saves 50% of developer time by translating natural language into pipeline commands, sandbox deployments, and build triggers
- 03RAG pipeline with semantic-boundary-aware chunking and sub-second vector similarity retrieval
- 04Conversation memory enabling multi-turn Q&A with follow-up question handling across both agents
- 05CLI companion built with Python and Typer automating 12+ pipeline commands for sandbox creation and deployment visibility