raqeem 0.2.2

رقيم — transcribe Arabic audio via Cohere's open ASR model, from the CLI or any language.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

# raqeem

The `raqeem` command (رقيم) — transcribe Arabic audio from the terminal, or from any
language, using
[`CohereLabs/cohere-transcribe-arabic-07-2026`](https://huggingface.co/CohereLabs/cohere-transcribe-arabic-07-2026),
the most accurate open-source Arabic speech-recognition model, Apache-2.0.

```bash
cargo install raqeem
```

A single static binary with no runtime — it loads no model weights. Inference is delegated
to an endpoint you choose:

```bash
export COHERE_API_KEY=...            # from dashboard.cohere.com/api-keys
raqeem voice_note.ogg --lang ar
```

Your own vLLM, no key and no rate limits:

```bash
raqeem clip.wav \
  --provider openai \
  --endpoint http://localhost:8000/v1/audio/transcriptions \
  --lang ar
```

`--format json` is what programs consume — verbatim text, an Arabic-normalized form, and
provenance:

```json
{
  "text": "الطماطم بـ ١٢٫٥ جنيه",
  "text_normalized": "الطماطم ب 12.5 جنيه",
  "provider": "cohere",
  "model": "cohere-transcribe-arabic-07-2026",
  "language": "ar"
}
```

That JSON contract is the language-agnostic API: shell out from Python, Node, Go, anything.
(Python has a native wheel instead — `pip install raqeem`.)

The Cohere-scoped `$COHERE_API_KEY` is deliberately **never** sent to a self-hosted
`--endpoint`; use `--api-key` or `$RAQEEM_API_KEY` for those.

Prebuilt binaries, full docs, and the library:
[github.com/SufficientDaikon/raqeem](https://github.com/SufficientDaikon/raqeem).

Apache-2.0 — same as the model. All accuracy credit belongs to Cohere Labs.