Skip to main content
Insight Flow is a workspace for research monitoring, continuous intelligence collection, and knowledge capture. You can either configure monitors manually or start by telling the Agent what topics, people, or directions you want to track, then let it produce an editable monitor draft. From there, the system uses your configured sources, models, and destinations to continuously collect, filter, summarize, generate reports, and publish results. The current product already covers a full end-to-end workflow: on the source side it supports RSS, site scraping, GitHub Trending, Hugging Face, and community or academic sources such as OpenAlex, PubMed, Europe PMC, Reddit, and X; on the execution side it runs scheduled monitors and generates daily briefs, weekly digests, research reports, and paper recommendations; on the model side it supports OpenAI-compatible LLM endpoints as well as Codex, with AI routing for different processing stages; and on the output side it can publish to Notion, Obsidian, or RSS while keeping run history, report archives, and manual republish support.

Quickstart

Start the local frontend, backend, database, and docs site.

Sources

Manage RSS, site scraping, GitHub Trending, Hugging Face, and more.

Monitors

Create monitors manually or with Agent-assisted drafts, then run them on demand.

API Reference

Explore resource groups under /api/v1/.

​
Core Capabilities

  • Multi-source collection across RSS, site scraping, GitHub Trending, Hugging Face, OpenAlex, PubMed, Europe PMC, Reddit, and X
  • Manual monitor setup plus Agent-assisted monitor draft generation
  • Report generation for daily briefs, weekly digests, research reports, and paper recommendations
  • Per-stage AI routing with OpenAI-compatible LLM endpoints and llm_codex
  • Output targets including Notion, Obsidian, and RSS, with runs, archives, and manual republish support

​
Typical Flow

1

Configure Providers

Fill in llm_openai or llm_codex settings and run connectivity tests.
2

Configure Destinations

Enable Notion, Obsidian, or RSS as output targets.
3

Prepare Sources

Create or sync sources and confirm categories, strategy, and test results.
4

Create a Monitor

Choose sources, report type, AI routing, and destinations, or start from an Agent-generated draft.
5

Review Results

Run the monitor and inspect reports, runs, logs, and feed output.

​
Next

  • Start with Quickstart if you want a working local setup
  • Read the Features pages to understand product workflows
  • Read the API Reference to integrate with the backend