langchain-ai/local-deep-researcher

Features

• Local Deep Research Agents - Provides a fully local deep research agent that iteratively researches topics using locally hosted LLMs.
• Autonomous Web Research Loops - Implements autonomous cycles of web search, summarization, and gap identification to deepen research.
• Agentic Workflow Engines - Provides a programmable engine for executing autonomous agent tasks with state persistence and checkpointing.
• Ollama Engine Integrations - Integrates with locally running Ollama and LMStudio engines for on-device AI research tasks.
• Ollama Research Agents - Provides a research agent that runs entirely on local hardware using Ollama-hosted LLMs.
• Local LLM Research Assistants - Provides a research assistant that uses locally hosted LLMs for web research and cited summaries.
• Cited Report Compilers - Compiles multi-source web findings into a final report with explicit citations for each source.
• Research Report Drafting - Iteratively retrieves and refines information to draft comprehensive research reports with citations.
• Search Query Recursion - Iteratively generates new search queries based on data gathered in previous research cycles.
• Stateful Agent Orchestrators - Manages agent state, tool sequencing, and persistence in multi-step research workflows.
• Research Automations - Conducts web research entirely offline using locally hosted LLMs without cloud API dependencies.
• Web Research Agents - Automates iterative web research by generating queries, summarizing findings, and identifying knowledge gaps.
• Local LLM Research Agents - Ships a research agent that runs entirely on local hardware using LM Studio-hosted LLMs.
• Iterative Research Tools - Provides a multi-step research loop that generates queries, summarizes results, and identifies knowledge gaps.
• Web Research Workflows - Automates the cycle of query generation, web search, summarization, and gap identification for thorough exploration.
• Multi-Source Report Compilers - Compiles findings from multiple web searches into a final report with citations for every source.
• Hybrid Short-and-Long Term Memory - Combines short-term working memory with persistent long-term storage across sessions.
• Agent Response Streams - Streams agent responses and tool results incrementally for real-time display and progressive processing.
• Incremental Result Streaming - Streams agent outputs incrementally to clients for real-time progress visibility.
• Durable Workflow Executions - Runs stateful, durable agent workflows that survive failures and resume from checkpoints.
• State Modifications - Pauses agent execution for human review and modification of internal state before resuming.
• Research Assistants - Local assistant for web research and writing.

Star history