garden-md

What is garden-md?

garden-md is a command-line tool that turns meeting transcripts (and related notes) into a structured, linked company knowledge base. It generates a Wikipedia-style HTML wiki from locally stored markdown files, so you can browse your organization’s people, companies, products/tools, and how they’re mentioned across documents.

Its core purpose is to help teams organize scattered transcript material into a browseable format while keeping the original transcripts verbatim (not rewritten or summarized). You can sync data from supported services, extract entities, link mentions, and open the resulting wiki in your browser.

Key Features

• Sync transcripts from multiple sources (local staging + local wiki): Runs garden sync to pull new transcripts from Grain, Granola, Fireflies.ai, or an API-based “Other” connector into a local staging area.
• Entity extraction across documents: During garden tend, the tool extracts entities such as people, companies, and products/tools and uses them to organize wiki pages.
• Linked, browsable wiki pages in markdown: The output is made of linked .md files that grow over time as more documents reference existing or new entities.
• Wikipedia-style HTML rendering: garden open generates and serves an HTML wiki with navigation, search, and backlinks.
• Local-first operation: The wiki is stored as folders of markdown files on your machine rather than in a cloud database.
• Optional git integration / automated commits: garden init can enable git, and the wiki can be updated with commits automatically (optional per setup).
• Connector approach with review for custom services: Built-in connectors are pre-built and tested; for “Other,” the AI generates a connector script shown for review before running.
• Clarified non-goals (preserves transcripts): It does not rewrite or summarize content; transcripts remain verbatim in the sourced documents.

How to Use garden-md

1. Install and prerequisites: Install with npm install -g garden-md. Use Node.js 18+ and provide an API key for a supported AI provider.
2. Initialize your wiki: Run garden init to choose your AI provider, configure the wiki location, and optionally enable git.
3. Connect a data source: Run garden connect and select a service—Grain, Granola, Fireflies.ai, or Other—to configure syncing. For “Other,” provide the service name and API documentation URL.
4. Sync new transcripts: Run garden sync to pull new transcripts into the staging area (shown as “wildland” in the demo output).
5. Process into the wiki: Run garden tend to extract entities, create stub pages, and generate a linked markdown/wiki structure; then run garden open to browse the generated HTML wiki.

Common commands include: garden connect --repair (to fix broken connectors), garden sync --schedule (cron-based automatic sync), and garden tend / garden open to process and browse.

Use Cases

• Turn meeting transcripts into a searchable internal wiki: After syncing transcripts from a meeting-note/transcript source, run garden tend to generate entity pages and backlinks so team members can browse company knowledge by person, company, or product/tool.
• Keep a “company brain” backed by verbatim source material: Because garden-md avoids rewriting/summarizing, teams can rely on the original transcript content while still benefiting from a structured wiki layer.
• Continuously update wiki pages as new meetings arrive: Use garden sync (optionally with --schedule) to bring in new transcripts, then re-run garden tend so the wiki grows and linkages expand as additional documents reference entities.
• Onboard teams or stakeholders with a common navigation layer: Use garden open to browse an HTML wiki with sidebar navigation and search, making it easier to find relevant mentions across many transcripts.
• Integrate a custom transcript source via API: Choose “Other” in garden connect and supply API documentation; the tool generates a connector script for review before running.

FAQ

• Does garden-md rewrite or summarize my transcripts? No. garden-md keeps transcripts verbatim (it does not rewrite or summarize content).

• Where is the wiki stored? As local markdown files (folders of .md files) on your machine; it does not require a database or cloud storage.

• Which services are supported out of the box? The repository lists built-in tested connectors for Grain, Granola, and Fireflies.ai. For other services, you can use “Other” with API documentation.

• What do garden tend and garden open do? garden tend reads from the staging area, sends content to your AI provider for entity extraction, and updates the linked markdown wiki. garden open renders and serves the HTML wiki for browsing.

• What prerequisites are required? The README states Node.js 18+ is required, and you need an API key from any supported AI provider.

Alternatives

• Build a wiki manually with markdown + scripts: If you only need basic organization, you can create and link markdown pages yourself without AI-driven entity extraction and HTML rendering workflows.
• Use a documentation/wiki platform with import pipelines: General-purpose knowledge bases can store and search documents, but the workflow differs because garden-md specifically generates Wikipedia-style cross-linked entity pages from transcripts.
• Transcript summarization tools with knowledge bases: Some tools focus on summarizing transcripts and indexing those summaries; garden-md’s approach is different because it preserves the original text and builds structured entity links rather than replacing content with summaries.
• Custom ETL pipelines (sync → parse → index): For teams that already have engineering support, you can replicate the “sync + entity extraction + linking” pattern using your own tooling and storage, instead of using garden-md’s command set and wiki renderer.