production AI platform kernel: LLM abstraction, workflow engine, RAG, MCP server, FastAPI, evaluation framework, closed-loop feedback
View on GitHub ↗The platform kernel for an AI content-automation system: one versioned, tested Python package that four content pipelines depend on, plus the service layer, persistence, observability, and an MCP server that turn those pipelines into a production platform.
It began as a way to eliminate ~1,500 lines of logic copy-pasted across four repositories. It has since grown into the shared foundation for LLM access, retrieval, agent orchestration, workflow execution, cost tracking, and a REST API — while keeping every addition justified by a real need (see docs/adr/).
Four interfaces, one engine — the CLI pipelines, the REST API, the MCP server, and the web dashboard all drive the same core. No business logic is duplicated across them.
Module Responsibility ------ contentcore.llm.LLMProvider Provider-agnostic text generation (Claude/OpenAI/Gemini), lazy SDK imports, retry + backoff, normalized responses, telemetry hook contentcore.workflow Step-based workflow engine — named stages, per-stage retry, optional-stage skipping, run reports contentcore.retry Retry decorator; retries only listed exception types (auth errors fail fast) contentcore.rag Semantic perspective retrieval — embeddings + cosine search, fingerprint-cached index, graceful fallback contentcore.agents Multi-agent story production — Director → Writer/Visual/Audio with a bounded single-revision critique loop contentcore.telemetry Per-call LLM token/cost/latency recording, structured JSON logging, optional Langfuse export contentcore.db SQLAlchemy persistence — workflow runs, step records, LLM usage, generated content c
From the project README.
Add the radar badge to your README — it shows your project was picked up by MCP Radar and links to this page:
[](https://mcp.liqiwa.com/s/sai-chaithanya-navuluri--content-core.html)
Self-hosted documentation RAG pipeline that serves semantic search to LLM agents over the Model Context Protocol.
SujalXplores/Agent-KCode-enforced incident-response agent that investigates SigNoz alerts, publishes evidence-backed root-cause claims, and gates every action behind a safety policy. Built for Agents of SigNoz hackathon.
bybit-exchange/kaasTurn scattered notes, docs and transcripts into a queryable Markdown wiki — an LLM knowledge-base compiler with MCP access, no embeddings, self-hosted.
vinodborole/okf-kitTurn any website into a portable, agent-ready Open Knowledge Format (OKF) bundle — no LLM required. Crawl, sync, chat offline, and serve docs to coding agents over MCP.
AlexanderBoger/SaveContextGit LFS for your LLM context — 20x Context Savings — an MCP server that keeps big documents out of the model and returns byte-exact quotes for a fraction of the tokens.
Alex663028/pulse-agentPulse — Hermes-style self-improving AI agent. Reliability-first rebuild with evaluated skill self-evolution, multi-agent team orchestration, dialectic user modeling, and fully self-hosted default stac
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.