Spaces:
Running
Running
File size: 2,749 Bytes
c7c5494 587c99f |
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 80 81 82 83 |
---
title: README
emoji: 📈
colorFrom: purple
colorTo: green
sdk: static
pinned: false
---
<a href="https://symbiont.dev" target="_blank"><img src="https://github.com/ThirdKeyAI/Symbiont/raw/main/logo-hz.png"></a>
# 🐙 Symbiont on Hugging Face
**Secure AI-agent runtime + specialized SLMs for high-trust workflows.**
We build small, safety-forward models and datasets that power **policy-gated agents**—with verifiable audit trails, strong privacy, and reproducible evaluation.
> 💡 Symbiont = an agent framework + DSL designed for zero-trust, cryptographically auditable automation. This org hosts the **models, datasets, and Spaces** that plug into Symbiont (and any standard HF/Transformers stack).
---
Main repo: https://github.com/ThirdKeyAI/Symbiont
Python SDK: https://github.com/ThirdKeyAI/symbiont-sdk-python
JS SDK: https://github.com/ThirdKeyAI/symbiont-sdk-js
Symbiont-Demos: https://github.com/ThirdKeyAI/symbiont-demos
---
## 🔧 What you’ll find here (COMING SOON)
* **Models (SLMs & classifiers)**
* Tool-use & routing specialists (reasoning-forward, small context).
* Safety & policy classifiers (e.g., tool-call SAFE/UNSAFE).
* Domain mini-experts (e.g., OSINT, compliance screening, data wrangling).
* **Datasets**
* **Tool-calling** JSONL corpora (inputs → tools → expected calls).
* **Safety** corpora (policy violations, redaction targets, PII/secret patterns).
* **Domain** sets (financial filings snippets, entity linking, sanctions lexicons).
* **Spaces (demos)**
* Policy-gated agent sandboxes (no secrets stored; ephemeral sessions).
* Evaluation dashboards (benchmarks, error taxonomies, confusion matrices).
---
## 🛡️ Principles
* **Zero-trust by default:** models are small, auditable, and wrapped in policy.
* **No training leakage:** org models don’t learn from your inputs.
* **Reproducibility:** fixed seeds, dataset versions, and training cards per release.
* **Privacy first:** redaction pipelines + opt-in logging with content digests only.
* **Cryptographic traceability:** signed artifacts and hash-chained audit logs.
---
## 📣 Stay in touch
* Website/Docs: [https://symbiont.dev](https://symbiont.dev) *(or your preferred site)*
* Email: [[email protected]](mailto:[email protected])
---
## 📜 Citation
If you use our models or datasets in research, please cite:
```
@software{symbiont_ai_agents,
title = {Symbiont: A Secure, Policy-Gated AI-Agent Runtime and SLM Suite},
year = {2025},
url = {https://symbiont.dev},
author = {ThirdKey.ai}
}
```
---
> Questions or special licensing needs (enterprise/redistribution)? Reach out—happy to help you deploy models inside **policy-enforced, auditable** agent workflows.
|