meetstream-migrate

One-command migration from Recall.ai to MeetStream.ai. Scans your codebase, maps every API call, shows the diff, and migrates — safely, with warnings for anything that needs a human look.

v2.1.2 — near feature parity with Recall.ai (verified live against production, July 2026). MeetStream matches Recall across recording, per-participant streams, real-time audio/video/ transcript, chat + image output, Google and Outlook calendar, scheduling, signed-in bots, and retention — plus native AI summaries and MIA voice agents that Recall doesn't have. The only gaps: pause/resume recording and live video output into the call (both shipping soon). See Feature parity & gaps below.

---

Quick Start

pip install meetstream-migrate
meetstream-migrate scan ./your-recall-project

Commands

# Scan (read-only) — finds every Recall.ai reference, writes NOTHING
meetstream-migrate scan ./your-project

# Dry run — shows exactly what would change, still writes nothing
meetstream-migrate migrate ./your-project --dry-run

# Migrate — applies all transformations (prompts before writing)
meetstream-migrate migrate ./your-project --yes

# Verify your MeetStream credentials + API reachability
meetstream-migrate test --api-key YOUR_MEETSTREAM_API_KEY

CLI reference

meetstream-migrate <command> [path] [options]

Commands:
  scan [path]        Scan only — show what would change
  migrate [path]     Scan & migrate a project
  test                Run API tests against MeetStream.ai

Options:
  --yes, -y          Skip the confirmation prompt
  --dry-run          Show changes without writing files
  --api-key KEY      MeetStream API key (for the test command)
  --verbose, -v      Detailed output
  --help, -h         Show help

What it does

1. Scan — recursively finds:

• Recall.ai base URLs (us-east-1.recall.ai, eu-west-1.recall.ai, api.recall.ai)
• API endpoint calls (/api/v1/bot/, /api/v1/bot/{id}/audio/, etc.)
• Recall-specific field names (meeting_url, recording_mode, transcription_options, metadata, …)
• Webhook handlers (bot.status_change, joining_call, in_call_recording, call_ended, done, fatal, …)
• SDK imports (require('recall-ai'), import recall)
• Environment variables (RECALL_API_KEY, RECALL_API_URL, RECALL_WEBHOOK_SECRET)

Supported file types: .js .ts .jsx .tsx .mjs .cjs .py .go .rb .java .kt .rs .php .cs .swift .env .yaml .yml .json .toml .ini .cfg

2. Display — endpoint mapping table, field mapping table, webhook event mapping, and a compatibility check (auto-migratable vs. needs manual review).

3. Transform — base URL swap, endpoint path rewrites, field renames, webhook event renames, env var renames, import/require updates. Every change carries an inline warning when it needs a second look.

4. Report — what changed, what needs manual review, and a post-migration checklist.

---

Endpoint Mapping

Recall.ai	MeetStream.ai	Notes
POST /api/v1/bot/	POST /api/v1/bots/create_bot	Field names differ (see below)
GET /api/v1/bot/{id}/	GET /api/v1/bots/{id}/detail	Also GET /api/v1/bots/{id}/status
POST /api/v1/bot/{id}/leave_call/	GET /api/v1/bots/{id}/remove_bot	⚠️ Method changes POST → GET
DELETE /api/v1/bot/{id}/	DELETE /api/v1/bots/{id}/delete
GET /api/v1/bot/	GET /api/v1/bots	List all bots
GET /bot/{id}/audio/	GET /bots/{id}/get_audio	Presigned S3 URL, valid 1h
GET /bot/{id}/video/	GET /bots/{id}/get_video	Presigned S3 URL, valid 10min
GET /bot/{id}/transcript/	GET /transcript/{transcript_id}/get_transcript	⚠️ Uses transcript_id, not bot_id
GET /bot/{id}/participants/	GET /bots/{id}/get_participants
GET /bot/{id}/chat_messages/	GET /bots/{id}/get_chats
POST /bot/{id}/send_chat_message/	POST /bots/{id}/send_message	REST body { "message": "..." } — drop Recall's to/pin fields
POST /bot/{id}/output_video/	POST /bots/{id}/send_image	Public img_url (not base64)
POST /bot/{id}/output_audio/	WebSocket sendaudio command	⚠️ Manual — no REST equivalent, use socket_connection_url bridge
POST /bot/{id}/output_media/	WebSocket bridge (partial)	⚠️ Manual — covers audio + static image today
POST /bot/{id}/pause_recording/	—	⛔ Not yet supported (shipping soon)
POST /bot/{id}/resume_recording/	—	⛔ Not yet supported (shipping soon)
GET /bot/{id}/screenshots/	GET /bots/{id}/get_screenshots
GET /bot/{id}/speaker_timeline/	GET /bots/{id}/get_speaker_timeline
POST /api/v1/webhook/	callback_url field on create_bot	Per-bot, not account-level
GET /bot/{id}/calendar/…	POST /calendar/create_calendar (underscore)	Body: google_client_id/google_client_secret/google_refresh_token

Field Mapping

Recall.ai	MeetStream.ai	Notes
meeting_url	meeting_link
recording_mode (string)	video_required (boolean)	"speaker_view"/"gallery_view" → true, "audio_only" → false
metadata	custom_attributes	Values must be strings
transcription_options	recording_config.transcript.provider.{provider}	Object structure differs — see provider config
noone_joined_timeout	voice_inactivity_timeout
destination_url	webhook_url (nested in live_transcription_required)
assembly_ai	assemblyai	Also: assemblyai uses speech_models[] + language_code, not model/language
output_media.websocket.url	socket_connection_url.websocket_url

Webhook Events — live-verified against production

MeetStream's envelope key is event — the same key Recall uses. No envelope rename needed. Alongside it: bot_id, message, status_code (200 success / 500 failure), custom_attributes, and on lifecycle events, bot_status.

{ "event": "bot.inmeeting", "bot_id": "...", "bot_status": "InMeeting",
  "message": "...", "status_code": 200, "custom_attributes": {} }

Two-layer model: bot.stopped fires once; bot_status tells you why — Stopped (normal) / NotAllowed (lobby timeout) / Denied (host denied) / Error (crash). Recall's separate end-reason events collapse into this single event + status.

Recall.ai Event	MeetStream event	bot_status	status_code
bot.joining_call	bot.joining (may fire up to 3×)	Joining	200
bot.in_waiting_room	bot.in_waiting_room	InWaitingRoom	200
bot.in_call_not_recording	bot.inmeeting	InMeeting	200
bot.in_call_recording	bot.recording	Recording	200
(none — new)	bot.leaving	Leaving	200
(none — new)	bot.error (non-terminal streaming-provider error; bot keeps running)	InMeeting	— (omitted)
bot.call_ended	bot.stopped	Stopped	200
bot.call_ended (lobby timeout)	bot.stopped	NotAllowed	500
bot.call_ended (host denied)	bot.stopped	Denied	500
bot.fatal	bot.stopped	Error	500
recording.done	manifest.completed → audio.processed → video.processed (if video_required)	—	200
transcript.done	transcription.processed	—	200
transcript.failed	transcription.failed	—	500
(post-call)	bot.done — post-call providers only	Done	200 / 500
(retention)	data_deletion	—	200

Streaming-only providers (deepgram_streaming, assemblyai_streaming, meeting_captions) do not fire transcription.processed/transcription.failed/bot.done — their terminal event is audio.processed. Don't wait on bot.done for a streaming bot, or your handler hangs forever.

transcript_id is never in the webhook payload. Resolve it via GET /bots/{id}/detail (bot_details.transcript_id), the create_bot response, or GET /bots/{id}/transcriptions.

Full bot_status enum: Joining, InWaitingRoom, InMeeting, Recording, Leaving, Stopped, NotAllowed, Denied, Error, Done.

Real-time events

Recall.ai	MeetStream	Notes
transcript.data / transcript.partial_data	delivered to live_transcription_required.webhook_url	Chunk shape: { speakerName, timestamp, transcript, words[] }
participant_events.join / .leave	same event names, requires recording_config.realtime_endpoints	Different payload shape — nested under data.bot.{id,metadata}, no top-level bot_id

---

Breaking changes to review manually

1. Transcript retrieval uses transcript_id (from the create_bot response, /detail, or /transcriptions), not bot_id.
2. recording_mode (Recall string) → video_required (MeetStream boolean).
3. Chat messages: Recall's separate POST /send_chat_message/ → MeetStream's POST /bots/{id}/send_message.
4. Webhook registration: Recall registers globally via POST /webhook/; MeetStream uses callback_url per bot at creation.
5. Webhook event values rename (see table above) — but the envelope key stays event, so most handler code needs a value rename, not a restructure.
6. Post-call processing arrives as discrete events (manifest.completed, audio.processed, transcription.processed/transcription.failed, video.processed, bot.done) instead of one bot.done-then-fetch signal.
7. Live transcription: real_time_transcription.destination_url → live_transcription_required.webhook_url on create_bot.
8. Pause/resume recording has no MeetStream equivalent yet — remove these calls, or segment via remove_bot + re-create.
9. Output audio/media into the call is REST on Recall, WebSocket-bridge on MeetStream — needs a real refactor, not a rename (flagged, not auto-migrated).

Post-Migration Checklist

1. Set MEETSTREAM_API_KEY (get one at app.meetstream.ai/api-keys)
2. Review every ⚠️ warning from the migration output
3. Update webhook handlers: keep branching on event, rename the values, and check bot_status for bot.stopped's reason
4. Store transcript_id for transcript retrieval — from create_bot, /detail, or /transcriptions
5. Add handlers for manifest.completed, audio.processed, transcription.processed/failed, video.processed
6. Verify the recording_mode → video_required boolean logic
7. Remove any pause_recording/resume_recording calls (not supported yet)
8. Run meetstream-migrate test --api-key YOUR_KEY to confirm connectivity

---

Feature parity & gaps

Recall capability	MeetStream	Migrator behaviour
Create/list/get/remove/delete bot	✅	auto-mapped
Recording (audio/video) + retention	✅	auto-mapped
Per-participant streams	✅	get_audio_streams / get_recording_streams
Real-time audio/video/transcript	✅	live_audio_required / live_video_required / live_transcription_required
Send chat message	✅	send_chat_message → REST send_message
Output image/camera frame	✅	output_video → send_image (public URL)
Calendar (Google + Outlook) + scheduling	✅	calendar/events/{id}/bot → calendar/schedule/{id}
Signed-in / OAuth bots	✅	Google login domains/logins; Zoom OBF
AI summaries	✅ (MeetStream-only)	GET /bots/{id}/summary
Conversational voice agents (MIA)	✅ (MeetStream-only)	agent_config_id on create_bot
Output audio / interactive media	⚠️ partial	WebSocket bridge (manual, flagged)
Pause/resume recording	⛔ not yet	flagged — shipping soon
Live video output into the call	⛔ not yet	send_image covers static frames only

Run scan first to see exactly which of these appear in your codebase before migrating.

---

Requirements

Python ≥ 3.8

Development

pip install -e .
python -m pytest test/    # or: python -m meetstream_migrate.cli test --api-key YOUR_KEY

License

MIT © MeetStream.ai — meetstream.ai · docs.meetstream.ai