truthcoder commited on
Commit
1778d84
·
verified ·
1 Parent(s): fa70939

Model card + release notes (replace stale test-release README)

Browse files
README.md ADDED
@@ -0,0 +1,77 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ library_name: mlc-llm
4
+ base_model:
5
+ - mlabonne/Qwen3-4B-abliterated
6
+ - mlabonne/Qwen3-1.7B-abliterated
7
+ - mlabonne/Qwen3-0.6B-abliterated
8
+ base_model_relation: quantized
9
+ pipeline_tag: text-generation
10
+ tags:
11
+ - webgpu
12
+ - webllm
13
+ - mlc-llm
14
+ - on-device
15
+ - offline
16
+ - private
17
+ - abliterated
18
+ - uncensored
19
+ - qwen3
20
+ - gguf
21
+ - ollama
22
+ - mobile
23
+ ---
24
+ # TC Mini — a private AI that runs entirely on your phone or PC
25
+
26
+ **Open-source, on-device chat AI. No account, no server, no telemetry — your chats never leave your device.** Runs in the browser with WebGPU (phones, laptops, desktops) or on your PC with Ollama / LM Studio / llama.cpp. Three sizes: **4B** (desktop), **1.7B** (balanced / phones), **0.6B** (any phone, 336 MB).
27
+
28
+ ▶ **Open it on your phone or computer:** https://truthprotectors.github.io/TC-Mini/app/
29
+ 💻 **Code (GitHub):** https://github.com/truthprotectors/TC-Mini
30
+
31
+ ## What's inside
32
+
33
+ | Build | Size (browser) | For | Model |
34
+ |---|---|---|---|
35
+ | Desktop · 4B | 2.3 GB | laptops / desktops with a GPU or Apple Silicon | Qwen3-4B abliterated |
36
+ | Balanced · 1.7B | 939 MB | newer phones, small laptops | Qwen3-1.7B abliterated |
37
+ | Mobile · 0.6B | 336 MB | any WebGPU phone | Qwen3-0.6B abliterated |
38
+
39
+ **Honest lineage:** these are the open **Qwen3** models with refusal directions removed ("abliterated", by [mlabonne](https://huggingface.co/mlabonne)), quantized by us to 4-bit for WebGPU (MLC `q4f16_1` / `q4f32_1`; the 4B uses AWQ-style channel scales folded into the norms, which cut 4-bit quality loss from KL 0.120 to 0.090 vs f16). They are not yet our own distilled model — each signed manifest says so. Measured: 0/8 refusals on our 8-prompt probe for every size; ~34 tok/s for the 4B in Chrome on an RTX 5070; the 0.6B is good for simple code and explanations, weak at arithmetic.
40
+
41
+ ## Built to be trusted
42
+
43
+ - **Signed releases.** Every release has an Ed25519-signed `MANIFEST.json` listing the SHA-256 of every file. The app pins the public key (`02c6bedb…77ae`), verifies the signature in the browser, then hashes every downloaded shard before using it. A tampered mirror is refused before a single byte loads. Rollbacks to older releases are refused too.
44
+ - **Your device decides the size.** A capability check (GPU, shader-f16, memory, iOS/Android) picks a build that won't freeze your device. Phones never get the 4B, and integrated GPUs get a small context and prefill chunks, because a 4B model can hard-freeze a laptop's display driver.
45
+ - **Checked on your device.** Before the first chat, the exact build runs an 8-prompt response check on your hardware.
46
+ - **No telemetry.** The only network requests are the signed release files. Everything else runs locally, and it works offline after the first download.
47
+
48
+ ## Use it on a PC (Ollama / LM Studio / llama.cpp)
49
+
50
+ GGUF files (Q4_K_M and Q8_0 for every size) are on the [Hugging Face page](https://huggingface.co/truthcoder/TC-Mini/tree/main/gguf).
51
+
52
+ ```bash
53
+ # Ollama
54
+ ollama run hf.co/truthcoder/TC-Mini:Q4_K_M # picks the 4B
55
+ # llama.cpp
56
+ llama-cli -hf truthcoder/TC-Mini:Q4_K_M
57
+ ```
58
+ LM Studio: search **truthcoder/TC-Mini**.
59
+
60
+ ## Verify a release yourself (any machine with OpenSSH ≥ 8.2)
61
+
62
+ ```bash
63
+ git clone https://huggingface.co/truthcoder/TC-Mini && cd TC-Mini/release/r1-2026-09-20-full
64
+ A=qwen3-4b-abl-q4f16_1
65
+ ssh-keygen -Y verify -f allowed_signers -I truthcoder-mini-release -n truthcoder-mini -s $A/MANIFEST.json.sig < $A/MANIFEST.json
66
+ python3 -c "import json;m=json.load(open('$A/MANIFEST.json'));[print(v['sha256'],' $A/'+k) for k,v in m['files'].items()]" | sha256sum -c
67
+ ```
68
+
69
+ ## Self-host
70
+
71
+ The app is static files — no build step. Serve `app/`, `vendor/`, `probe.js`, `fixtures.json` and `probe_constants.json` from any web server. To serve the weights yourself, put the `release/` folder from Hugging Face next to `app/` and set `window.TC_WEIGHTS=""` in `app/boot.js`. Any mirror is safe because of the signatures.
72
+
73
+ ## License
74
+
75
+ App code: Apache-2.0. Model weights: Apache-2.0, inherited from [Qwen3](https://huggingface.co/Qwen) via [mlabonne/Qwen3-4B-abliterated](https://huggingface.co/mlabonne/Qwen3-4B-abliterated) (and the 1.7B / 0.6B siblings). Runtime: [WebLLM / MLC-LLM](https://github.com/mlc-ai/web-llm) (Apache-2.0).
76
+
77
+ Abliterated models answer requests that stock models refuse. You are responsible for how you use it.
release/r1-2026-09-20-full/README.md CHANGED
@@ -1,7 +1,11 @@
1
- # truthcoder-mini release `test-2026-09-15` — TEST (placeholder weights, test key)
 
 
 
 
2
  Verify any artefact on any machine with OpenSSH ≥ 8.2:
3
  ```
4
- ssh-keygen -Y verify -f allowed_signers -I truthcoder-mini-test@truthstack.app -n truthcoder-mini -s <artefact>/MANIFEST.json.sig < <artefact>/MANIFEST.json
5
- sha256sum -c <(python3 -c "import json;m=json.load(open('<artefact>/MANIFEST.json'));[print(v['sha256'],' ','<artefact>/'+k) for k,v in m['files'].items()]")
6
  ```
7
- Get it: the `.torrent`/magnet in RELEASE.json (BitTorrent), or the CID in RELEASE.json (IPFS: `ipfs get <cid>`), or any listing mirror — then verify as above. The signature binds the manifest; the manifest binds every byte.
 
1
+ # TC Mini release r1-2026-09-20 (signed)
2
+
3
+ Stock abliterated Qwen3 weights (mlabonne), quantized by us for WebGPU (MLC). Release sequence 2.
4
+ Signer: `truthcoder-mini-release`, Ed25519 key SHA256:lLR0sZXfFRlUHOwmqj65J+iG3JaLMgeeANEhpYLP4RQ (raw pubkey hex 02c6bedb56aceafb46bd8053055e1d8e67420db115cda0d4df3c20076a1c77ae).
5
+
6
  Verify any artefact on any machine with OpenSSH ≥ 8.2:
7
  ```
8
+ ssh-keygen -Y verify -f allowed_signers -I truthcoder-mini-release -n truthcoder-mini -s <artefact>/MANIFEST.json.sig < <artefact>/MANIFEST.json
9
+ python3 -c "import json;m=json.load(open('<artefact>/MANIFEST.json'));[print(v['sha256'],' <artefact>/'+k) for k,v in m['files'].items()]" | sha256sum -c
10
  ```
11
+ `MANIFEST.json.ed25519` is the raw Ed25519 signature the browser app checks. Also available by BitTorrent (the `.torrent` files here).
release/r1-2026-09-20-mid/README.md CHANGED
@@ -1,7 +1,11 @@
1
- # truthcoder-mini release `test-2026-09-16-mid` — TEST (placeholder weights, test key)
 
 
 
 
2
  Verify any artefact on any machine with OpenSSH ≥ 8.2:
3
  ```
4
- ssh-keygen -Y verify -f allowed_signers -I truthcoder-mini-test -n truthcoder-mini -s <artefact>/MANIFEST.json.sig < <artefact>/MANIFEST.json
5
- sha256sum -c <(python3 -c "import json;m=json.load(open('<artefact>/MANIFEST.json'));[print(v['sha256'],' ','<artefact>/'+k) for k,v in m['files'].items()]")
6
  ```
7
- Get it: the `.torrent`/magnet in RELEASE.json (BitTorrent), or the CID in RELEASE.json (IPFS: `ipfs get <cid>`), or any listing mirror — then verify as above. The signature binds the manifest; the manifest binds every byte.
 
1
+ # TC Mini release r1-2026-09-20 (signed)
2
+
3
+ Stock abliterated Qwen3 weights (mlabonne), quantized by us for WebGPU (MLC). Release sequence 2.
4
+ Signer: `truthcoder-mini-release`, Ed25519 key SHA256:lLR0sZXfFRlUHOwmqj65J+iG3JaLMgeeANEhpYLP4RQ (raw pubkey hex 02c6bedb56aceafb46bd8053055e1d8e67420db115cda0d4df3c20076a1c77ae).
5
+
6
  Verify any artefact on any machine with OpenSSH ≥ 8.2:
7
  ```
8
+ ssh-keygen -Y verify -f allowed_signers -I truthcoder-mini-release -n truthcoder-mini -s <artefact>/MANIFEST.json.sig < <artefact>/MANIFEST.json
9
+ python3 -c "import json;m=json.load(open('<artefact>/MANIFEST.json'));[print(v['sha256'],' <artefact>/'+k) for k,v in m['files'].items()]" | sha256sum -c
10
  ```
11
+ `MANIFEST.json.ed25519` is the raw Ed25519 signature the browser app checks. Also available by BitTorrent (the `.torrent` files here).
release/r1-2026-09-20-mobile/README.md CHANGED
@@ -1,7 +1,11 @@
1
- # truthcoder-mini release `test-2026-09-16-mobile` — TEST (placeholder weights, test key)
 
 
 
 
2
  Verify any artefact on any machine with OpenSSH ≥ 8.2:
3
  ```
4
- ssh-keygen -Y verify -f allowed_signers -I truthcoder-mini-test -n truthcoder-mini -s <artefact>/MANIFEST.json.sig < <artefact>/MANIFEST.json
5
- sha256sum -c <(python3 -c "import json;m=json.load(open('<artefact>/MANIFEST.json'));[print(v['sha256'],' ','<artefact>/'+k) for k,v in m['files'].items()]")
6
  ```
7
- Get it: the `.torrent`/magnet in RELEASE.json (BitTorrent), or the CID in RELEASE.json (IPFS: `ipfs get <cid>`), or any listing mirror — then verify as above. The signature binds the manifest; the manifest binds every byte.
 
1
+ # TC Mini release r1-2026-09-20 (signed)
2
+
3
+ Stock abliterated Qwen3 weights (mlabonne), quantized by us for WebGPU (MLC). Release sequence 2.
4
+ Signer: `truthcoder-mini-release`, Ed25519 key SHA256:lLR0sZXfFRlUHOwmqj65J+iG3JaLMgeeANEhpYLP4RQ (raw pubkey hex 02c6bedb56aceafb46bd8053055e1d8e67420db115cda0d4df3c20076a1c77ae).
5
+
6
  Verify any artefact on any machine with OpenSSH ≥ 8.2:
7
  ```
8
+ ssh-keygen -Y verify -f allowed_signers -I truthcoder-mini-release -n truthcoder-mini -s <artefact>/MANIFEST.json.sig < <artefact>/MANIFEST.json
9
+ python3 -c "import json;m=json.load(open('<artefact>/MANIFEST.json'));[print(v['sha256'],' <artefact>/'+k) for k,v in m['files'].items()]" | sha256sum -c
10
  ```
11
+ `MANIFEST.json.ed25519` is the raw Ed25519 signature the browser app checks. Also available by BitTorrent (the `.torrent` files here).