Migrate the Speechify API to Cartesia

Move this codebase from Speechify API to Cartesia for text to speech. Keep Speechify API working behind a flag until I switch over.

November 21, 2026: Simba 3.5 launch pricing ends. Simba 3.5 bills at $10, $8 and $6 per million characters on Starter, Pro and Scale until then, and at its list price of $20, $16 and $12 after. The Free plan's Simba 3.5 allowance drops from 500,000 to 250,000 characters a month on the same day. (https://speechify.ai/pricing)

November 21, 2026: Simba 1.6 voices change under the old model ids. `simba-english` and `simba-multilingual` stopped being selectable from API version `2026-09-21`. Workspaces pinned to an older version keep the Simba 1.6 audio only until November 21, when current models start serving both ids and the audio changes. (https://docs.speechify.ai/docs/models)

## 1. Find the calls
Search for `api.speechify.ai` or `SPEECHIFY_API_KEY`. The request to replace is `POST https://api.speechify.ai/v1/audio/speech`. List every call site before you edit anything.

## 2. Target
- Request: `POST https://api.cartesia.ai/tts/bytes`
- Model: `model_id: "sonic-3.6"`
- Headers: `X-API-Key` read from the `CARTESIA_API_KEY` environment variable, and `Cartesia-Version: 2026-08-14`
- Auth swap: Speechify API sends the key in `Authorization: Bearer <key>` from `SPEECHIFY_API_KEY`. Replace it with `X-API-Key` from `CARTESIA_API_KEY` on the Cartesia path only, and keep `SPEECHIFY_API_KEY` for the Speechify API path. Keys come from the console at platform.speechify.ai, and the Speechify SDKs read `SPEECHIFY_API_KEY` on their own.
- Add `CARTESIA_API_KEY` wherever the project documents its environment (`.env.example`, README, deploy config). Never hardcode the key.
- Request size: Speechify API caps a request at 2,000 spoken characters on `/v1/audio/speech`; 20,000 on `/v1/audio/stream`. Cartesia publishes no per-request maximum, so keep any chunking the code already does and split at sentence boundaries; for long text, stream over the Cartesia TTS WebSocket.
- Concurrency: Speechify API allows 3 concurrent requests on Free, 15 on Starter, 30 on Pro, 60 on Scale, shared across every synthesis endpoint. Look up the Cartesia plan's concurrent request limit at https://www.cartesia.ai/pricing, cap parallel requests at it, and retry rate-limit errors with backoff.
- Voice IDs don't carry over. Keep each voice ID in one named constant or config value, put a placeholder where each Speechify API voice was, and list them for me so I can pick replacements at https://www.cartesia.ai/voices.
- Suggested starting points to put in those placeholders, for me to confirm by ear: A female voice for articles and long-form reading → `a33f7a4c-100f-41cf-a1fd-5822e8fc253f` (Lauren - Lively Narrator); A male voice for chapters and documentaries → `ed82c17b-4704-4d34-be43-5d19065acdf1` (Carl - Steady Storyteller); A conversational voice for an in-app assistant → `f786b574-daa5-4673-aa0c-cbe3e8534c02` (Katie - Friendly Fixer).

## 3. Field mapping, Speechify API → Cartesia
- `POST https://api.speechify.ai/v1/audio/speech` → `POST https://api.cartesia.ai/tts/bytes`
- `POST /v1/audio/stream` (chunked bytes) → The TTS WebSocket. Sonic sends audio chunks while it generates; the [WebSocket reference](https://docs.cartesia.ai/api-reference/tts/websocket) has the message format.
- `Authorization: Bearer sk_...` → `X-API-Key` and `Cartesia-Version: 2026-08-14`
- `input`, text or SSML, up to 2,000 spoken characters → `transcript`. Cartesia reads only the `<break>`, `<spell>`, `<emotion>`, `<speed>` and `<volume>` tags. Any other SSML has no effect.
- `voice_id` → `voice`: `{"id": "<voice id>"}`
- `model`: `simba-3.0` (default), `simba-3.2` or `simba-3.5` → `model_id`: `sonic-3.6`. `simba-english` and `simba-multilingual` are retired. A workspace on API version `2026-09-21` or later gets `400 model_retired` for either id.
- `language`, such as `en-US` → `language`: `en` or `en-US`
- `audio_format`: `wav`, `mp3`, `pcm` → `output_format.container` and `encoding`. `pcm` is the `raw` container with `pcm_s16le`.
- `output_format`: `ulaw_8000` → `raw` container, `pcm_mulaw` encoding, `sample_rate` 8000
- `pronunciations`, up to 200 per request and outside English for now → `pronunciation_dict_id`. Copy the entries into one Cartesia dictionary and pass its ID.
- `speech_marks` in the response → `add_timestamps` on the TTS WebSocket. `/tts/bytes` returns no timings. Word timestamps arrive as WebSocket events.
- Response `audio_data` (base64) → Response body is the audio bytes

## 4. Speechify API fields with no Cartesia equivalent
- `audio_format`: `ogg`, `aac`. What to do: Ask for `mp3` or `wav` and convert it yourself.

## 5. Reference request
Match the language and HTTP client the codebase already uses. This is the same request in Python:
```python
import os
import requests

text = "Chapter four begins on a cold morning in late November."
VOICE_ID = "6ccbfb76-1fc6-48f7-b71d-91ac6298247b"  # from /voices

response = requests.post(
    "https://api.cartesia.ai/tts/bytes",
    headers={
        "X-API-Key": os.environ["CARTESIA_API_KEY"],
        "Cartesia-Version": "2026-08-14",
    },
    json={
        "model_id": "sonic-3.6",
        "transcript": text,
        "voice": {"id": VOICE_ID},
        "language": "en",
        "output_format": {
            "container": "wav",
            "encoding": "pcm_s16le",
            "sample_rate": 44100,
        },
    },
)
response.raise_for_status()
open("out.wav", "wb").write(response.content)
```

## 6. Roll out
1. Choose Speechify API or Cartesia per request from one config flag, defaulting to Speechify API.
2. Run the same scripts through both and save the clips side by side for me to listen to. Check that the container, encoding and sample rate match what downstream code expects.
3. Leave the Speechify API client, call and key in place so rollback is flipping the flag. Don't delete them in this change.
4. Run the existing tests, and add one that checks the Cartesia request body.

## Docs
- Speechify audio speech reference: https://docs.speechify.ai/api-reference/audio/speech
- Speechify audio stream reference: https://docs.speechify.ai/api-reference/audio/stream
- Speechify authentication: https://docs.speechify.ai/build/guides/get-started/authentication
- Cartesia TTS bytes API reference: https://docs.cartesia.ai/api-reference/tts/bytes

When you're done, tell me which files changed, which fields you couldn't map, and how to flip the flag.