Uploaded September 2026 | Updated November 2026, 4 days ago
Build a production grade RAG system in LangChain with Qdrant, hybrid search, metadata filters, and cited answers so the LLM stops hallucinating when the document does not contain the answer.
This is the final lesson of the LangChain 10 Days Bootcamp, and it pulls everything together. You will ingest Amazon and Google 10-K filings into a Qdrant vector database using RAGWire with a single YAML config, see how content hashing blocks duplicate ingestion, run manual retrieval with rag.retrieve, get grounded answers with citations from rag.query, watch an out of scope question trigger the refusal rule, and then build your own agent with create_agent that uses get_filter_context and search_documents tools, a model call limit, and a checkpointer for follow-up questions. Hybrid search (dense + sparse, BM25, reciprocal rank fusion) is explained on the whiteboard. Memory, tools, agents, and chunking were covered in earlier days, so if you are new, start with the playlist.
📺 Full playlist: youtube.com/playlist?list=PLW4pPr9JCovI
⏱ Chapters:
0:00 Intro and what we cover today
1:37 What is RAG: Retrieve, Augment, Generate
2:43 RAGWire library and the YAML config idea
4:21 Ingestion vs retrieval
5:56 Hybrid search: dense vs sparse, BM25, RRF
8:04 Install RAGWire and uv sync
9:38 Start Qdrant with Docker Compose
11:16 The dataset: Amazon and Google 10-K filings
12:53 .env keys and LangSmith tracing
13:57 Connect the Qdrant client
14:59 RAG config YAML walkthrough
18:08 Metadata extraction and dedup chunks
20:15 Create the collection and ingest the PDFs
22:19 Where the metadata fields come from
22:55 Content hash blocks duplicate ingestion
24:30 rag.retrieve: manual retrieval
25:32 rag.query: answers with citations
27:13 Tracing the run in LangSmith
28:17 Hallucination on an out of scope question
30:57 The grounding rule and refusal
33:01 Controlling top_k
34:35 Breaking complex questions into sub-queries
35:43 Comparison answer and citation fields
38:22 answer.formatted()
39:53 Custom agent: get_filter_context and search_documents tools
42:03 create_agent with middleware and checkpointer
43:40 Run the agent and inspect the messages
45:46 Follow-up question with memory
47:21 Adding citations to your own agent and wrap-up
🔗 Resources:
Code + slides: github.com/laxmimerit/Langchain-10-Days-Bootcamp
RAGWire install: pip install "ragwire[all]" (or just run uv sync in the repo)
Free models on OpenRouter: openrouter.ai/models
LangChain install docs: docs.langchain.com/oss/python/langchain/install
UV package manager setup: youtube.com/watch?v=pcYaP1lGd3s&list=PLc2rvfiptPSQdX4jeLgW4nf5PId4j1msI&index=8
📺 Watch first (Day 9, vector DB and ingestion): youtube.com/watch?v=qXOchjCKt0A&list=PLW4pPr9JCovI
📺 Watch next (agents with memory): youtube.com/watch?v=O0Cj4nef21Q&list=PLW4pPr9JCovI
🎓 Go deeper - my Udemy courses:
Master LangChain v1 and Ollama - Chatbot, RAG and AI Agents: kgptalkie.com/langchain
Advanced RAG: Build & Deploy Production GenAI Apps: kgptalkie.com/advanced-rag
If this video helped, a like goes a long way. Comment where you got stuck, especially with Docker or Qdrant, and subscribe with the bell on so you catch the next playlist from KGP Talkie. Thanks for sticking with all ten days.
#LangChain #RAG #Qdrant #GenerativeAI #AIAgents
Build a production grade RAG system in LangChain with Qdrant, hybrid search, metadata filters, and cited answers so the LLM stops hallucinating when the document does not contain the answer.
This is the final lesson of the LangChain 10 Days Bootcamp, and it pulls everything together. You will ingest Amazon and Google 10-K filings into a Qdrant vector database using RAGWire with a single YAML config, see how content hashing blocks duplicate ingestion, run manual retrieval with rag.retrieve, get grounded answers with citations from rag.query, watch an out of scope question trigger the refusal rule, and then build your own agent with create_agent that uses get_filter_context and search_documents tools, a model call limit, and a checkpointer for follow-up questions. Hybrid search (dense + sparse, BM25, reciprocal rank fusion) is explained on the whiteboard. Memory, tools, agents, and chunking were covered in earlier days, so if you are new, start with the playlist.
📺 Full playlist: youtube.com/playlist?list=PLW4pPr9JCovI
⏱ Chapters:
0:00 Intro and what we cover today
1:37 What is RAG: Retrieve, Augment, Generate
2:43 RAGWire library and the YAML config idea
4:21 Ingestion vs retrieval
5:56 Hybrid search: dense vs sparse, BM25, RRF
8:04 Install RAGWire and uv sync
9:38 Start Qdrant with Docker Compose
11:16 The dataset: Amazon and Google 10-K filings
12:53 .env keys and LangSmith tracing
13:57 Connect the Qdrant client
14:59 RAG config YAML walkthrough
18:08 Metadata extraction and dedup chunks
20:15 Create the collection and ingest the PDFs
22:19 Where the metadata fields come from
22:55 Content hash blocks duplicate ingestion
24:30 rag.retrieve: manual retrieval
25:32 rag.query: answers with citations
27:13 Tracing the run in LangSmith
28:17 Hallucination on an out of scope question
30:57 The grounding rule and refusal
33:01 Controlling top_k
34:35 Breaking complex questions into sub-queries
35:43 Comparison answer and citation fields
38:22 answer.formatted()
39:53 Custom agent: get_filter_context and search_documents tools
42:03 create_agent with middleware and checkpointer
43:40 Run the agent and inspect the messages
45:46 Follow-up question with memory
47:21 Adding citations to your own agent and wrap-up
🔗 Resources:
Code + slides: github.com/laxmimerit/Langchain-10-Days-Bootcamp
RAGWire install: pip install "ragwire[all]" (or just run uv sync in the repo)
Free models on OpenRouter: openrouter.ai/models
LangChain install docs: docs.langchain.com/oss/python/langchain/install
UV package manager setup: youtube.com/watch?v=pcYaP1lGd3s&list=PLc2rvfiptPSQdX4jeLgW4nf5PId4j1msI&index=8
📺 Watch first (Day 9, vector DB and ingestion): youtube.com/watch?v=qXOchjCKt0A&list=PLW4pPr9JCovI
📺 Watch next (agents with memory): youtube.com/watch?v=O0Cj4nef21Q&list=PLW4pPr9JCovI
🎓 Go deeper - my Udemy courses:
Master LangChain v1 and Ollama - Chatbot, RAG and AI Agents: kgptalkie.com/langchain
Advanced RAG: Build & Deploy Production GenAI Apps: kgptalkie.com/advanced-rag
If this video helped, a like goes a long way. Comment where you got stuck, especially with Docker or Qdrant, and subscribe with the bell on so you catch the next playlist from KGP Talkie. Thanks for sticking with all ten days.
#LangChain #RAG #Qdrant #GenerativeAI #AIAgents










