Spaces:
Running
Running
File size: 2,147 Bytes
607bb36 2801c36 607bb36 2801c36 a7ecac7 8aaa576 a7ecac7 8aaa576 0e92ba7 451445c 8aaa576 2801c36 a7ecac7 0e92ba7 8aaa576 2801c36 a7ecac7 0e92ba7 2801c36 8aaa576 a7ecac7 0e92ba7 | 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 69 70 71 72 73 74 75 76 77 78 79 | ---
title: PRISM-Memory
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 6.12.0
app_file: app.py
pinned: false
---
# PRISM-Memory Space
**Hook:** Turn conversations into durable, searchable memory.
This Space is the public demo for the released `PRISM-Memory 7B Adapter`, a
LoRA adapter on top of `Qwen/Qwen2.5-7B-Instruct`.
It matches the root repo story:
- one released model
- one extraction behavior
- one set of confirmed benchmark results
- one interactive turn-by-turn demo
- one compact explanation of the synthetic training data and held-out examples
Within the GitHub repo, this directory is a first-class public surface. The
helper script `scripts/deploy_space.sh` bundles this folder with the release
artifacts needed for Hugging Face.
## Inputs
The app reads:
- `results/release_summary.json`
- `results/extraction_examples.json`
- `results/try_it_sessions.json`
- `docs/release/extraction-skill.md`
- `docs/release/datasets.md`
When copied into a standalone Hugging Face Space repo, keep those files beside
`app.py` and `requirements.txt`, preserving the `docs/release/` and `results/`
subdirectories.
## What It Shows
1. The confirmed metrics for the released model
2. A `Try It` flow with bundled multi-turn sessions and pasted transcript input
3. Side-by-side held-out extraction examples against the GPT-4.1 reference
4. A compact description of the synthetic conversation data and supervision
5. The single canonical memory extraction skill to keep
The bundled example sessions use precomputed outputs from the released model.
Custom pasted sessions use live model inference when the runtime can load the
adapter; otherwise the Space falls back to a lightweight contract-preserving
preview.
## Local Run
```bash
python -m pip install -r requirements.txt
python app.py
```
## From The Repo Root
```bash
bash scripts/deploy_space.sh
```
Useful bundled files:
- `docs/release/extraction-skill.md`
- `docs/release/datasets.md`
- `docs/release/extraction-examples.md`
- `docs/release/release-results.md`
- `results/release_summary.json`
- `results/extraction_examples.json`
- `results/try_it_sessions.json`
|