📡 OPENROUTER
MODEL REGISTRY

OMR polls OpenRouter's full model catalog on a schedule, detects every change — new models, removals, pricing shifts, and free-status transitions — and proposes a ranked fallback chain of validated free models for the platform's AI apps. By default, a probe-validated chain goes live automatically the moment it's proposed — flip one setting if you'd rather review and approve each chain by hand first.

Premier Active v1.0.0 Auth Required
4
Change Dimensions
:free
Tier Tracked
Chain
Proposed Ordering
Auto
Approve (Default)

Everything Needed to Keep the Fallback Chain Current

🔄
Scheduled Catalog Polling
A cron-driven cycle pulls OpenRouter's complete model list and snapshots it to omr_snapshots / omr_models. Each run captures the full catalog state for diffing.
🔍
4-Dimension Change Detection
Diffs each snapshot against the last to flag: new models added, models removed, pricing changes, and models that gained or lost :free status — all logged with timestamps.
🧪
Free-Model Validation
Candidate free models are probed for liveness before entering the chain. Only models that pass validation are considered for the proposed fallback order.
📋
Ranked Chain Proposals
Validated free models are scored and ordered into a proposed fallback chain (A3 first), logged to omr_chain. Every proposal is versioned — full history available.
🔔
Telegram Alerts
Material changes — models gaining or losing free status, a new proposed chain — trigger an immediate Telegram alert with a summary of what changed and why it matters.
Auto-Approve, With a Manual Gate on Standby
By default a probe-passing chain is promoted to active the same cycle it's proposed — no click needed. Set OMR_AUTO_APPROVE=false to require an explicit dashboard Approve before anything goes live instead.
💬
Free Model Chat
Multi-turn chat with any free model in the catalog. Conversation history is persisted in the database and replayed as full context on every send — the model remembers previous turns. The header shows the selected model's context window, price, and when OMR first saw it; a card highlights the newest free model, and a live meter tracks context usage as you chat.
🥊
Model Arena
Send one prompt to multiple free models simultaneously and compare every response side by side. Click any result to open a slide panel with full model details and continue the conversation with that specific model — without leaving the screen.

Keeping the AI Fallback Chain Current, Automatically

OMR runs a fully automated detect → propose → alert → (auto-)approve loop. Every step is logged; by default the last step happens automatically, with an optional switch to require human sign-off on every chain change instead.

01
Poll the catalog
A scheduled cycle pulls OpenRouter's full model list. Every model's ID, name, pricing, and context window is captured and stored in omr_snapshots / omr_models.
02
Detect changes
The new snapshot is diffed against the previous one. Changes are classified into four buckets: new model, removed, pricing change, free-status shift.
03
Propose the fallback chain
Validated free models are probed for liveness and ordered into a proposed chain (A3 first). The proposal is logged to omr_chain with its full ordered list and reasoning.
04
Alert on material changes
When the free-model set or proposed chain shifts, OMR fires a Telegram alert summarising what became free, what lost free status, and the full proposed chain — before you even open the dashboard.
05
Goes live automatically (or on your approval)
By default, a probe-passing chain is promoted to active the same cycle — downstream AI apps (A3 and others) pick it up immediately, no code change or redeploy. Prefer to review first? Set OMR_AUTO_APPROVE=false and the chain waits in the dashboard for a one-click Approve instead.

POLL. DETECT. APPROVE. COMPARE.

Keep the platform's AI fallback chain accurate without manual catalog checks.
Every free-model gain or loss surfaces automatically. Chat and compare models directly.

Premier app — requires MelTuc account