Nagovori
Back to blog
5 minNagovori

Build a Knowledge Base from Calls Using Obsidian, Claude Code, and Nagovori

ObsidianClaude Codeknowledge-baseautomation

Build a Knowledge Base from Calls Using Obsidian, Claude Code, and Nagovori

Every week, your team has meetings. Decisions are made, context is shared, plans are discussed. A week later, nobody remembers the specifics. Two months later, the same discussions repeat because institutional knowledge lives in people's heads, not in a system.

This guide shows you how to build a searchable, interconnected knowledge base from your recorded calls — using Nagovori for transcription, Claude Code for intelligent structuring, and Obsidian for long-term storage.

Why This Stack

  • Nagovori — converts audio to text fast and accurately. Upload a file, get clean text in minutes.
  • Claude Code — an AI coding assistant that runs locally and can process text files, create structured notes, and automate repetitive formatting. It works with your filesystem directly, which makes it perfect for Obsidian vault manipulation.
  • Obsidian — a Markdown-based knowledge manager with bidirectional links, graph view, and full-text search. Your notes are plain .md files — no vendor lock-in.

The Workflow

Step 1: Record and Transcribe

Record your meeting (phone, Zoom, Google Meet — any format works). After the call, upload the recording to Nagovori. A one-hour meeting typically processes in 2–3 minutes. Download or copy the transcript.

Step 2: Save the Raw Transcript

Create a folder structure in your Obsidian vault:

/Calls
  /2026-04-19 — Product sync
    transcript.md
    summary.md
  /2026-04-18 — Client onboarding
    transcript.md
    summary.md

Paste the raw transcript into transcript.md. Add basic metadata in the YAML frontmatter:

---
date: 2026-04-19
type: meeting
participants: [Alice, Bob, Carol]
project: product-launch
tags: [sync, decisions]
---

Step 3: Generate Structured Notes with Claude Code

This is where Claude Code shines. Open your terminal in the Obsidian vault directory and ask Claude Code to process the transcript:

claude "Read Calls/2026-04-19 — Product sync/transcript.md and create
a summary.md in the same folder with:
1. Meeting context (one sentence)
2. Key decisions made (bulleted list)
3. Action items with owners and deadlines
4. Open questions
5. Links to related topics using [[wiki-links]]
Also, if any topic maps to an existing note in my vault, add
backlinks."

Claude Code reads the transcript, understands the content, and generates a well-structured summary with Obsidian-compatible [[wiki-links]]. It can also scan your existing vault to create connections to prior notes.

Step 4: Build Connections Over Time

After a few weeks, your knowledge graph starts forming naturally:

  • [[Product Launch]] gets mentioned in five different meeting summaries
  • [[Client Feedback]] links to three call transcripts and two decision notes
  • [[API Migration]] connects engineering syncs with client-facing updates

Obsidian's graph view shows these connections visually. You can trace how a decision evolved across meetings.

Step 5: Automate with a Script

If you do this daily, automate the boring parts. Create a simple shell script:

#!/bin/bash
# Process today's recordings
DATE=$(date +%Y-%m-%d)
VAULT="$HOME/ObsidianVault/Calls"

for file in ~/Downloads/*.m4a ~/Downloads/*.mp3; do
  [ -f "$file" ] || continue
  NAME=$(basename "$file" | sed 's/\.[^.]*$//')
  DIR="$VAULT/$DATE — $NAME"
  mkdir -p "$DIR"
  # Upload to Nagovori (manual step — or use the Telegram bot)
  # After transcription, save transcript.md and run Claude Code
  echo "Process: $DIR"
done

For the transcription step, you can forward recordings to Nagovori's Telegram bot and copy the results, or upload through the web interface.

Advanced Patterns

Weekly Digest

Ask Claude Code to generate a weekly summary across all meeting notes:

claude "Read all summary.md files in Calls/ from this week.
Create a Weekly Digest note with: top decisions, all open action items
(grouped by owner), and recurring themes."

Project Knowledge Pages

Create a living document for each project that auto-updates with references:

claude "Find all meeting summaries that mention [[Product Launch]].
Update the Product Launch note with a timeline of decisions and
current status based on the latest meetings."

Search by Person

With proper participants frontmatter, you can query: "What did Alice commit to in the last 30 days?" Claude Code can search across all meeting notes and compile a report.

What This Costs

  • Nagovori: ~60 ₽ per hour of audio (at 1.4 ₽/min package rate) — roughly $0.65
  • Claude Code: included in your Anthropic subscription
  • Obsidian: free for personal use

For a team that has 5 hours of meetings per week, the transcription costs about $3.25/week. The time saved on note-taking and information retrieval pays for itself after the first week.

Tips

  • Tag consistently. Use the same tags across meetings: #decision, #action-item, #blocker. This makes filtering trivial.
  • Don't over-structure. Start simple. Let the knowledge graph emerge from natural linking rather than imposing a rigid hierarchy upfront.
  • Review weekly. Spend 15 minutes each Friday scanning the week's summaries. Close completed action items, update project pages.
  • Keep raw transcripts. Never delete them — they're your source of truth when summaries miss nuance.

Conclusion

The combination of fast transcription (Nagovori), intelligent structuring (Claude Code), and a flexible knowledge manager (Obsidian) creates a system where institutional knowledge accumulates automatically. Every meeting becomes a searchable, linkable artifact instead of a fading memory.