Spaces:
Sleeping
Sleeping
| # data/static/ | |
| Static identity files for ArunCore. These define who Arun is and how the agent is allowed to behave. They are the foundation layer of the knowledge base — loaded on every query regardless of what is retrieved. | |
| --- | |
| ## Files | |
| ### `public_profile.md` | |
| Arun's public-facing professional identity. Contains: | |
| - Name, title, and one-line positioning | |
| - Core engineering stack | |
| - Key projects (names + one-line descriptions) | |
| - Professional values and working style | |
| - What he's currently building and looking for | |
| This file is injected as baseline context for every agent response. It ensures the agent always has a minimum accurate representation of Arun even for questions that don't match any specific project document. | |
| ### `rules_of_engagement.md` | |
| The agent's answer policy. Defines: | |
| - Truth-only rule: never speculate or fill gaps with plausible-sounding information | |
| - Source citation requirement: every factual claim must be traceable | |
| - Fallback behavior: what the agent says when it doesn't have the information | |
| - Tone guidelines: professional, direct, first-person | |
| --- | |
| ## How These Files Are Used | |
| When a query arrives, both files are included in the LLM context window as fixed system-level context. Retrieved chunks from the vector DB are appended on top. The agent synthesises an answer grounded in both layers. | |
| --- | |
| ## Update Frequency | |
| These files should be updated whenever Arun's professional positioning, stack, or answer policy changes. They do not need to be re-embedded — they are injected directly as text, not retrieved via vector search. | |