docs: restructure README — screenshots, slim comparison

[saucam]

Why

The README was 699 lines: a 20-row comparison matrix dominated the top, the product screenshots (already on yashdatta.dev) were missing, and the whole feature reference lived inline. This makes it a scannable overview that links out for depth — closer to how top OSS projects lead.

What changed

Slimmed the README 699 → 257 lines. Front page is now: hero → why → compare → quick start → architecture → condensed Features overview → interfaces (screenshots) → short pointers → CLI/dev. Detail moved to docs:

• docs/FEATURES.md — full feature reference (terminal keybindings, slash commands, memory ranking, context-reduction layers, Telegram commands, production-resilience table).
• docs/CONFIGURATION.md — permission scopes + every env var + config.json schema + ~/.codeoid/.env.
• docs/COMPARISON.md — the full feature-by-feature matrix (was inline).

Screenshots (were missing):

• Hero is now the flagship Rust TUI cockpit; web UI + Telegram shots live in the new Interfaces section; the architecture diagram moved into the Architecture section (it was the hero).

Accuracy fix (fact-checked against the real omnigent source):

• The old matrix wrongly claimed Omnigent has no cross-session recall — it has search_conversations + a long-term memory store. Reframed as memory shape (workspace-scoped verbatim episodes vs conversation FTS), not memory presence. Dropped the unverifiable "open-sourced by Databricks" line; bumped Omnigent's mid-turn-input cell to ✅.

Verification

• All images and doc links resolve; relative links that moved into docs/ were re-pointed (../src/…, cross-doc anchors); no dangling #terminal-client / #web-ui hero anchors.

🤖 Generated with Claude Code

---

Update: multi-harness accuracy pass

Caught while auditing: the README still called codeoid a single-harness (Claude-only) tool and contrasted that with Omnigent's breadth. That predates the provider registry — codeoid now ships six backends behind one SessionProvider interface (claude default, openai, gemini, codex, pi, gemini-cli), selectable per session and forkable across backends. Fixed the hero/architecture/features/footer, added a docs/FEATURES.md#harnesses section, and reframed the Omnigent comparison from breadth-vs-depth to memory/identity-vs-isolation (both are multi-harness). Matrix rows corrected: Multi-harness ❌→✅, Multi-model routing ❌ roadmap → ~.

[github-actions]

🤖 Gemini code review

The PR successfully restructures the README by relocating the large comparison table to a separate document (docs/COMPARISON.md), positioning the architecture diagram under its dedicated section, and adding professional screenshots for the Rust TUI, Web UI, and Telegram interfaces. Additionally, it refines and corrects the accuracy of the feature comparison regarding Omnigent's capabilities.

Findings: 🔴 0 · 🟠 0 · 🟡 0 · 🟢 0

---

_{Tokens spent · ⬆️ Input: 4,738 · ⬇️ Output: 86 · Σ Total: 7,778
Total may be higher due to thinking token counts.}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 84.90%. Comparing base (f4139b7) to head (d1acd66).
⚠️ Report is 2 commits behind head on main.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #176   +/-   ##
=======================================
  Coverage   84.90%   84.90%           
=======================================
  Files         106      106           
  Lines       18678    18678           
=======================================
  Hits        15858    15858           
  Misses       2820     2820           

Flag	Coverage Δ
daemon	84.90% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.