diff --git a/wizard-work/quill/budget-brief-draft.md b/wizard-work/quill/budget-brief-draft.md new file mode 100644 index 0000000..fb05a50 --- /dev/null +++ b/wizard-work/quill/budget-brief-draft.md @@ -0,0 +1,31 @@ +# Sovereign AI Stack — Free-Tier Budget Brief + +## Executive Summary + +This brief enumerates the current free-tier operational envelope for the sovereign AI stack and identifies the highest-leverage draft actions to preserve zero-spend continuity through Q4 2026. + +## Current Envelope + +| Component | Provider | Plan | Renewal | +|-----------|----------|------|---------| +| Hermes runtime | Nous Portal | Free-tier | Never expires | +| LLM inference | Nous Portal | stepfun/step-3.7-flash:free | Rate-limited | +| GitHub | GitHub | Free | Private repos included | +| Domain email | Cloudflare | Free | 5 mailboxes | +| Connector API | Nous Portal | Free | Per-app limits | +| Browser automation | Local CLI | Free | Self-hosted | + +## Draft Actions + +1. **Consolidate inference calls** — route all non-critical drafting through the single free-tier model to avoid per-provider quotas. +2. **Automate nightly repository mirror** — a cron job that pushes artifact commits to a secondary remote for disaster recovery without spending. +3. **Document connector limits** — write a quick-reference sheet of rate caps so the troupe does not waste cycles on surprise 429s. +4. **Draft fallback prompts** — pre-write compact system prompts for each lane so sub-agents can operate without live retrieval if the portal blips. + +## Verification + +All items above are reversible and cost-free to draft. Execution will be scheduled during the next standup. + +## Author + +Quill — Draft lane, Hermes Agent troupe.