π Build a production-grade document ingestion pipeline for RAG systems.
βοΈ Using @ApacheAirflow@FastAPI@PostgreSQL
π Learn DAG orchestration, idempotency, deduplication, status tracking, and reliable PDF processing workflows.
π§ Self-host @langfuse locally
π Track prompts, traces, latency & token usage
β‘ Connect everything to @vllm_project with @OpenAI compatible APIs
π³ Run the full observability stack with @Docker Compose
π§ Learn how Kimi-K2 stabilizes trillion-parameter LLM training with MuonClip QK-Clip
β‘ Built using DeepSeek-V3-style MoE MLA components
π οΈ Full PyTorch implementation included