Open-source connectors for Embedded — the AI-powered voice memo app. Record, transcribe, summarize, and search your voice memos with AI. Export your transcriptions, summaries, tags, and embeddings to the tools you already use.
git clone https://github.com/mattlsmith/embedded-connect.git
cd embedded-connect
pip install -r requirements.txtpython connectors/obsidian/export.py --email you@example.com --vault-path ~/Obsidian/Vault/VoiceMemospython connectors/json-export/export.py --email you@example.com --output my_memos.jsonfrom embedded import EmbeddedClient
client = EmbeddedClient()
client.login("you@example.com", "your-password")
memos = client.get_memos()
for memo in memos:
print(memo["category"], memo["summary"][:80])All you need is your Embedded account email and password — the same credentials you use in the iOS app. No API keys or tokens to manage.
Your data is fetched through a secure API that verifies your identity and returns only your memos. Other users' data is never accessible.
| Connector | Description | Path |
|---|---|---|
| Obsidian (Simple) | Markdown files with YAML frontmatter, organized by month | connectors/obsidian/export.py |
| Obsidian (Smart Vault) | Full vault system with dashboards, person routing, 1:1 files | connectors/obsidian/smart-vault/ |
| Notion | Database pages with properties, rich content, collapsible transcripts | connectors/notion/ |
| JSON Export | Full data dump as JSON — great for custom integrations | connectors/json-export/ |
Syncs your voice memos to a Notion database. Each memo becomes a page with structured properties and rich content.
-
Create a Notion integration at notion.so/my-integrations
-
Create a database in Notion with these columns:
Column Type Purpose Title Title Memo title (auto-created) Category Select Meeting, Idea, ToDo, etc. Tags Multi-select Your custom tags Date Date Memo creation date Memo ID Rich text For deduplication -
Share the database with your integration (click
···→ Connections → add your integration) -
Copy the database ID from the URL (the 32-char hex string)
# Set credentials (or pass as CLI flags)
export NOTION_TOKEN=secret_xxx
export NOTION_DATABASE_ID=abc123def456...
# Dry run
python connectors/notion/export.py --email you@example.com --dry-run
# Export all memos
python connectors/notion/export.py --email you@example.com
# Incremental sync
python connectors/notion/export.py --email you@example.com --incremental
# Filter by category
python connectors/notion/export.py --email you@example.com --category MeetingEach Notion page includes:
- Properties: Title, Category (select), Tags (multi-select), Date, Memo ID
- Callout block with metadata (category, chunk count, audio file)
- Summary section (for ToDo memos, action items become interactive checkboxes)
- Collapsible transcription (toggle block — click to expand)
- Deduplication — never creates duplicate pages
- Rate limit handling — automatic retry with backoff
A complete Obsidian knowledge management system that automatically routes your voice memos to the right place, manages 1:1 files for your team, and provides live dashboards.
- Auto-routing — Meetings go to person 1:1 files, ideas to an Ideas folder, ToDos to your inbox with checkboxes
- Person detection — Mentions of people in your memos are matched to their 1:1 file
- 1:1 file management — Each person gets a structured file with Action Items, Meeting Notes (newest-first), and Rapport sections
- Dataview dashboards — Action Dashboard, 1:1 Dashboard, and Rapport Dashboard that auto-aggregate from your vault
- File normalization — Person files maintain canonical section ordering, safe to run repeatedly
- Incremental sync — Only fetches new memos since last run
- Deduplication — Never ingests the same memo twice
# 1. Set up your vault structure
cd connectors/obsidian/smart-vault
python setup.py
# 2. Edit people.yaml in your vault — add your team
# 3. Install the Dataview plugin in Obsidian
# 4. Dry run (preview routing without writing)
python ingest.py --email you@example.com --dry-run
# 5. Execute
python ingest.py --email you@example.com --execute
# 6. Incremental sync (subsequent runs)
python ingest.py --email you@example.com --execute --incrementalyour-vault/
├── 00_Inbox/ ← ToDo items with checkboxes
├── 01_Dashboards/
│ ├── Action Dashboard ← Aggregates all open tasks
│ ├── 1on1 Dashboard ← Recent meetings, all people
│ └── Rapport Dashboard ← Personal details reference
├── 02_Voice_Memos/
│ ├── Meetings/ ← Standalone meeting files
│ └── General/ ← Other/uncategorized memos
├── 03_People/
│ └── {Team}/
│ └── Person Name.md ← 1:1 file (auto-created)
├── 04_Resources/
│ ├── Ideas/ ← Idea memos (article-spark format)
│ └── Budget/ ← Budget/forecast notes
├── 99_Templates/
│ └── Template - 1on1 Person.md
└── people.yaml ← Your team directory
| Category | Destination | Behavior |
|---|---|---|
| Meeting / People | Person 1:1 file | Appends to ## Meeting Notes if one person detected |
| Meeting / People | 02_Voice_Memos/Meetings/ |
Creates standalone file if no person or multiple people |
| Idea | 04_Resources/Ideas/ |
Article-spark format with summary + transcription |
| Budget | 04_Resources/Budget/ |
Budget note format |
| ToDo | 00_Inbox/ |
Extracts action items as - [ ] checkboxes |
| Other | 02_Voice_Memos/General/ |
General memo format |
Define your team so the pipeline can route memos to the right person:
me:
name: "Your Name"
people:
- name: "Jane Doe"
nickname: ["jane"]
team: "Engineering"
role: "Senior Engineer"
vault_path: "03_People/Engineering/Jane Doe.md"
- name: "Bob Wilson"
nickname: ["bob"]
team: "Design"
role: "Lead Designer"
vault_path: "03_People/Design/Bob Wilson.md"Customize folder paths, routing, and behavior:
vault_path: "/path/to/your/vault"
folders:
inbox: "00_Inbox"
meetings: "02_Voice_Memos/Meetings"
general: "02_Voice_Memos/General"
people: "03_People"
ideas: "04_Resources/Ideas"
budget: "04_Resources/Budget"
max_action_items: 10--email EMAIL Your Embedded account email (required)
--execute Write files (default is dry run)
--incremental Only fetch memos since last run
--category CATEGORY Filter: Meeting, Idea, ToDo, People, Budget, Other
--memo-id ID Process a single memo by ID
The shared client that all connectors use. You can also use it directly.
| Method | Description |
|---|---|
login(email, password) |
Authenticate with your Embedded account |
get_memos(since, category, include_embeddings) |
Fetch memos with stitched transcriptions |
get_raw_chunks(since, category, include_embeddings) |
Fetch raw embedding chunks (for search/ML) |
Every memo returned by get_memos() includes:
| Field | Type | Description |
|---|---|---|
memo_id |
string | Unique identifier |
transcription |
string | Full transcription (stitched from chunks) |
summary |
string | AI-generated executive summary |
category |
string | Meeting, Idea, ToDo, People, Budget, Other |
tags |
string | Comma-separated user tags |
audio_file_name |
string | Original audio file path |
created_at |
string | ISO 8601 timestamp |
updated_at |
string | ISO 8601 timestamp |
chunk_count |
int | Number of embedding chunks |
Pass include_embeddings=True to get the 3072-dimensional Gemini embedding vectors for each chunk. Useful for building semantic search, clustering, or recommendation features.
memos = client.get_memos(include_embeddings=True)
for memo in memos:
for chunk in memo["chunks"]:
vector = chunk["embedding"] # 3072-dim list of floats
print(f"Chunk {chunk['chunk_index']}: {len(vector)} dimensions")Use get_raw_chunks() when you need per-chunk data without stitching:
chunks = client.get_raw_chunks(include_embeddings=True)
for chunk in chunks:
print(chunk["memo_id"], chunk["chunk_index"], chunk["start_char"], chunk["end_char"])--email EMAIL Your Embedded account email (required)
--vault-path PATH Obsidian vault directory for memos (required)
--incremental Only export memos created since last run
--category CATEGORY Filter: Meeting, Idea, ToDo, People, Budget, Other
--dry-run Preview without writing files
--email EMAIL Your Embedded account email (required)
--output PATH Output JSON file path (required)
--category CATEGORY Filter by category
--since TIMESTAMP ISO timestamp — only memos after this date
--include-embeddings Include 3072-dim embedding vectors
--raw-chunks Export raw chunks instead of stitched memos
Create a new directory under connectors/ and use the shared client:
from embedded import EmbeddedClient
client = EmbeddedClient()
client.login(email, password)
# Get stitched memos (transcription chunks combined)
memos = client.get_memos()
# Or get raw chunks for embedding-level access
chunks = client.get_raw_chunks(include_embeddings=True)
# Build your integration here!See examples/ for more patterns.
We welcome community connectors! To add a new one:
- Fork this repo
- Create
connectors/your-connector/export.py - Use
embedded.pyfor auth and data fetching - Add a section to this README
- Open a pull request
Ideas for connectors:
- Logseq — export as Logseq-compatible markdown
- Roam Research — daily notes format
- CSV — spreadsheet-friendly export
- Todoist / Things — extract ToDo items as tasks
- Slack — post summaries to a channel
- Webhook — push new memos to any URL
- SQLite — local searchable database
embedded-connect/
├── embedded.py # Shared Python SDK
├── requirements.txt # Python dependencies
├── connectors/
│ ├── obsidian/
│ │ ├── export.py # Simple Obsidian export
│ │ └── smart-vault/ # Full vault system
│ │ ├── setup.py # Interactive setup wizard
│ │ ├── ingest.py # Smart routing pipeline
│ │ ├── normalize.py # Person file normalizer
│ │ ├── config.yaml # Pipeline configuration
│ │ └── vault-template/ # Starter vault structure
│ ├── notion/
│ │ └── export.py # Notion database connector
│ └── json-export/
│ └── export.py # JSON export connector
└── examples/
├── basic_usage.py # Fetch and print memos
└── search_memos.py # Keyword search across memos
- Your password is used only to authenticate with Firebase and is never stored or transmitted elsewhere
- The API verifies your identity server-side and returns only your data
- No API keys or service credentials are required
- Row Level Security (RLS) enforces per-user data isolation at the database level
- All communication is over HTTPS
MIT — see LICENSE.