File size: 1,834 Bytes
5b76cb0
 
 
 
 
 
 
 
 
e8e08ec
 
5b76cb0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9cdcce8
5b76cb0
 
3839eee
5b76cb0
7b406f4
fb59416
5b76cb0
 
 
 
 
 
 
9cdcce8
5b76cb0
 
 
 
3839eee
5b76cb0
7b406f4
fb59416
5b76cb0
9cdcce8
5b76cb0
 
 
9cdcce8
9d7051c
 
ab74a45
 
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
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
# Deployment

## Hugging Face Space

- Space repo: `https://huggingface.co/spaces/KPrashanth/agent-tina`
- Live app: `https://kprashanth-agent-tina.hf.space/`
- SDK: Docker
- App port: `7860`
- Runtime status last verified: `RUNNING` on 2026-06-08
- Dataset repo: `https://huggingface.co/datasets/KPrashanth/agent-tina-meetings`
- End-to-end meeting creation, transcription, correction, MoM generation, downloads, and Dataset persistence verified on 2026-06-08

## Local Deployment Script

Run:

```powershell
.\.venv\Scripts\python.exe scripts\deploy_hf_space.py
```

The script:

- reads credentials from `.env`
- creates or updates the Space
- uploads project files
- ignores `.env`, `.venv`, caches, bytecode, and `sample.wav`
- configures Space variables/secrets

## Required `.env` Values

```powershell
HF_TOKEN=...
HF_SPACE_NAME=agent-tina
HF_APP_BASE_URL=https://kprashanth-agent-tina.hf.space
HF_DATASET_REPO=KPrashanth/agent-tina-meetings
OPENROUTER_API_KEY=...
OPENROUTER_MODEL=openai/gpt-oss-20b
OPENROUTER_CORRECTION_MODEL=openai/gpt-4o-mini
WHISPER_MODEL=base
WHISPER_BEAM_SIZE=5
MAX_UPLOAD_MB=250
```

## Space Secrets and Variables

Secret:

- `OPENROUTER_API_KEY`
- `HF_TOKEN`

Variables:

- `OPENROUTER_MODEL`
- `OPENROUTER_CORRECTION_MODEL`
- `WHISPER_MODEL`
- `WHISPER_BEAM_SIZE`
- `MAX_UPLOAD_MB`
- `APP_BASE_URL`
- `HF_DATASET_REPO`

## Important Constraint

Hugging Face Space disk may be ephemeral unless persistent storage is enabled. The app optionally syncs meeting folders to `HF_DATASET_REPO` when `HF_TOKEN` and `HF_DATASET_REPO` are configured.

After a Space rebuild, meeting metadata is restored from the private Dataset on demand. Large original recordings and Markdown outputs are restored lazily only when requested.

Post-rebuild metadata and original-audio recovery were verified on 2026-06-09.