savant-bot v0.0.2 — Ship openrouter/free default fix

savant-bot v0.0.2 — Patch release

Ships the openrouter/auto → openrouter/free default-model fix from v0.0.1. No code or API changes; only default value, documentation, and version bumps. Wire-compatible with v0.0.1.

Fixed

• Default LLM model (src/config.rs): llm_default_model default corrected from openrouter/auto to openrouter/free. The earlier default was a wrong-interpretation bug (operator's original URL had openrouter/free as the literal model slug, not a navigation hint).
• README.md intro line: Updated to reference openrouter/free (was openrouter/auto).

Audit

• Verified unchanged: all src/*.rs source files, .env.example, migrations. ECHO Protocol conformity still v0.1.3 (this release bumps app version only).
• Validated: all 6 ECHO validation commands PASS; 25/25 unit tests pass.
• FID-151 N/A: no new pub fn introduced this release.
• Spec consistency: Cargo.toml v0.0.2 ↔ protocol.config.yaml project.version v0.0.2 ↔ CHANGELOG.md ↔ README.md v0.0.2.
• Tracked by: FID-2026-0616-011 (archived in dev/fids/archive/).

Upgrade path

�ash git pull origin main cargo build --release

Users who set LLM_DEFAULT_MODEL explicitly are unaffected. Users relying on the v0.0.1 default now get the corrected openrouter/free.

Dual-versioning note

This release bumps the app version (�.0.2). The protocol version (used for ECHO Protocol conformity tracking via VERSION + protocol.config.yaml:protocol.version) remains �.1.3 — the protocol itself did not change this release. See CHANGELOG.md preamble for the rule.

Full Changelog: �0.0.1...v0.0.2

SAVANT-BOT v0.0.1 - Initial Public Release

SAVANT-BOT v0.0.1 — Initial Public Release

Sister project to SAVANT-TRADING under the SAVANT umbrella, built under SAVANT-PROTOCOL (ECHO) v0.1.3.

Highlights

• Rust-native Discord bot on Poise 0.6 + Serenity 0.12 + Tokio
• 3 commands: /ping (smoke test), /ask (LLM chat), /mute (moderation, requires MODERATE_MEMBERS)
• LLM features (OpenRouter-backed): defer-then-edit, token-bucket rate limiting, exponential backoff on 429, sliding-window context
• Moderation features: SQLite poller for temp-punishments (restart-survival)
• 25 unit tests, all 6 validation commands PASS
• ~1,800 lines of Rust across 18 files in src/ + 1 SQL migration

Implementation (10 FIDs)

All FIDs implemented and archived to dev/fids/archive/. Highlights:

• FID-008 (Provider): Provider trait + OpenRouterProvider + MockProvider for tests. Uses X-OpenRouter-Title header per dev/docs/openrouter-llms.md.
• FID-005 (Defer): defer_and_run helper using Poise 0.6's ctx.defer() + ctx.say() pattern.
• FID-006 (Rate Limit): governor token-bucket (GCRA) + exponential backoff with ±20% jitter.
• FID-007 (Context): Sliding-window LLM context, 20 messages/channel/1h TTL.
• FID-004 (Poller): SQLite moderation_cases + 60s background poller.
• FID-010 (Skeleton): Cargo baseline + Poise framework setup.

Audit Results (release-time)

• All 6 validation commands PASS: cargo build, cargo check, cargo clippy --all-targets -- -D warnings, cargo fmt --check, cargo test (25/25), cargo clean
• Zero unwrap() / expect() in non-test code
• Zero TODO / FIXME / todo! / unimplemented! in src/
• Zero secrets in tracked files (.env properly gitignored)
• All pub fn wired to consumers (FID-151 call-graph reachability verified)

Configuration (verified at release)

Field	Value	Source
project.version	0.0.1	Cargo.toml, protocol.config.yaml
protocol.version	0.1.3	matches VERSION file
language	rust	config
quality.max_file_lines	300	coding-standards/rust.md
quality.max_function_lines	50	coding-standards/rust.md
quality.max_line_length	100	coding-standards/rust.md
perfection_loop.max_iterations	10	ECHO.md Rule #5
perfection_loop.change_threshold	0.10	ECHO.md Rule #1
perfection_loop.convergence_threshold	0.02	ECHO.md Rule #3
session.autonomy_level	3 (Autonomous)	ECHO default

Known Limitations → v1.0

• /mute records the case but does NOT call the Discord API to actually apply the mute (schema has role_id ready for v2)
• LLM context is in-process only; lost on bot restart (SQLite persistence is v1.1)
• Process-wide rate limiter (per-user keyed limiting is v1.1)
• No streaming SSE for long LLM responses (v2)

Roadmap (per README)

• v1.1: /mute actually mutes, context persistence, more moderation commands, per-user rate limiting
• v2.0: Full SAVANT memory integration (CortexaDB), soul/personality system, A2A delegation, multi-server sharding, voice channel support, custom skills

See Also

• README
• ECHO.md (SAVANT-PROTOCOL spec)
• CHANGELOG.md
• Sister project: SAVANT-TRADING
• Parent framework: SAVANT