Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Create and manage routing combos with 14 strategies (priority, weighted, round-robin, Auto-combo, etc.). Configure fallback chains, test routing outcomes, and retrieve combo metrics.
.claude/skills/diegosouzapw-omni-combos-routing/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 85% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 178% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 173% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 122% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 88% | 0% |
<!-- generated by src/lib/agentSkills/generator.ts; manual edits will be overwritten -->
Create and manage routing combos with 19 strategies (priority, weighted, round-robin, Auto-combo, and more). Configure fallback chains, test routing outcomes, and retrieve combo metrics.
All requests require a valid Bearer token or session cookie. Obtain a token via POST /api/auth/login or configure REQUIRE_API_KEY=false for local development.
List routing combos
bashcurl https://localhost:20128/api/combos \ -H "Authorization: Bearer $OMNIROUTE_TOKEN"
Create routing combo
bashcurl -X POST https://localhost:20128/api/combos \ -H "Authorization: Bearer $OMNIROUTE_TOKEN" \ -H "Content-Type: application/json" \ -d '{}'
Get combo by ID
bashcurl https://localhost:20128/api/combos/{id} \ -H "Authorization: Bearer $OMNIROUTE_TOKEN"
Update combo
Partial update: the body is merged onto the stored combo, so a field left out keeps its current value. An array that IS sent replaces the stored one outright.
bashcurl -X PUT https://localhost:20128/api/combos/{id} \ -H "Authorization: Bearer $OMNIROUTE_TOKEN" \ -H "Content-Type: application/json" \ -d '{}'
Update combo
Partial update: the body is merged onto the stored combo, so a field left out keeps its current value. An array that IS sent replaces the stored one outright.
bashcurl -X PATCH https://localhost:20128/api/combos/{id} \ -H "Authorization: Bearer $OMNIROUTE_TOKEN" \ -H "Content-Type: application/json" \ -d '{}'
Delete combo
bashcurl -X DELETE https://localhost:20128/api/combos/{id} \ -H "Authorization: Bearer $OMNIROUTE_TOKEN"
Get combo metrics
bashcurl https://localhost:20128/api/combos/metrics \ -H "Authorization: Bearer $OMNIROUTE_TOKEN"
Test a combo configuration
bashcurl -X POST https://localhost:20128/api/combos/test \ -H "Authorization: Bearer $OMNIROUTE_TOKEN" \ -H "Content-Type: application/json" \ -d '{}'
List fallback chains
Returns all registered fallback chains for model routing.
bashcurl https://localhost:20128/api/fallback/chains \ -H "Authorization: Bearer $OMNIROUTE_TOKEN"
Create fallback chain
Registers a fallback routing chain for a model.
bashcurl -X POST https://localhost:20128/api/fallback/chains \ -H "Authorization: Bearer $OMNIROUTE_TOKEN" \ -H "Content-Type: application/json" \ -d '{}'
Delete fallback chain
bashcurl -X DELETE https://localhost:20128/api/fallback/chains \ -H "Authorization: Bearer $OMNIROUTE_TOKEN"
See the full OpenAPI specification at GET /api/openapi/spec or docs/openapi.yaml for detailed request/response schemas.
<!-- skill:custom-start --> <!-- Migrated from skills/omniroute-routing/SKILL.md (preserved curated content) -->
Requires OMNIROUTE_URL and OMNIROUTE_KEY. See entry-point SKILL for setup.
A combo is a named group of providers/models with a routing strategy. All requests through a combo are automatically distributed, failed-over, and load-balanced — the caller uses a single model ID like my-combo.
bashcurl $OMNIROUTE_URL/api/combos \ -H "Authorization: Bearer $OMNIROUTE_KEY"
Response includes id, name, strategy, enabled, and per-target stats.
bashcurl -X POST $OMNIROUTE_URL/api/combos \ -H "Authorization: Bearer $OMNIROUTE_KEY" \ -H "Content-Type: application/json" \ -d '{ "name": "my-combo", "strategy": "priority", "targets": [ { "provider": "anthropic", "model": "claude-opus-4-7", "weight": 1 }, { "provider": "openai", "model": "gpt-4o", "weight": 1 } ] }'
| Strategy | Description | | --- | --- | | priority | Always use target0]; fall back on error | | weighted | Distribute by weight percentage | | round-robin | Rotate targets in order | | context-relay | Chain models for very long contexts | | fill-first | Fill quota of target0] before spilling | | p2c | Power-of-two choices: sample two targets, pick the better candidate | | random | Uniform random selection | | least-used | Route to the target with the lowest observed usage | | cost-optimized | Pick the cheapest eligible target for the token estimate | | reset-aware | Prefer targets based on quota-reset state | | reset-window | Order targets by their configured reset window | | headroom | Prefer targets with more remaining quota headroom | | strict-random | Random without repeating until all targets have been used | | auto | Auto-Combo scoring across 13 factors | | lkgp | Last-known-good-provider sticky routing | | context-optimized | Pick the best model for the request's context size | | cache-optimized | Prefer targets with stronger cache affinity | | fusion | Run a panel in parallel and synthesize one judged response | | pipeline | Run a configured sequence of targets as a pipeline |
Auto-combo scores each candidate on 13 factors every request:
bashcurl -X POST $OMNIROUTE_URL/api/combos \ -H "Authorization: Bearer $OMNIROUTE_KEY" \ -H "Content-Type: application/json" \ -d '{ "name": "prod-auto", "strategy": "auto", "targets": [ { "provider": "anthropic", "model": "claude-sonnet-4-6" }, { "provider": "openai", "model": "gpt-4o-mini" }, { "provider": "google", "model": "gemini-2.0-flash" } ] }'
Then call it with:
bashcurl -X POST $OMNIROUTE_URL/v1/chat/completions \ -H "Authorization: Bearer $OMNIROUTE_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "prod-auto", "messages": [{ "role": "user", "content": "Hello" }] }'
bash# Activate curl -X PUT $OMNIROUTE_URL/api/combos/{id}/toggle \ -H "Authorization: Bearer $OMNIROUTE_KEY" \ -d '{ "enabled": true }'
bashcurl $OMNIROUTE_URL/api/combos/{id}/metrics \ -H "Authorization: Bearer $OMNIROUTE_KEY"
Returns p50/p95/p99 latency, success rate, cost, and per-target breakdown.
bashcurl -X POST $OMNIROUTE_URL/api/routing/simulate \ -H "Authorization: Bearer $OMNIROUTE_KEY" \ -H "Content-Type: application/json" \ -d '{ "comboId": "{id}", "messages": [{ "role": "user", "content": "test" }] }'
Returns which provider would be selected and why — no actual API call is made.
omniroute_list_combos → list all combos
omniroute_switch_combo → enable/disable a combo
omniroute_set_routing_strategy → change strategy at runtime
omniroute_simulate_route → dry-run routing decision
omniroute_best_combo_for_task → get recommendation by task type404 combo not found → check id from /api/combos400 invalid strategy → use one of the 19 strategies above409 name conflict → combo name already exists<!-- skill:custom-end -->
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 7,956 | 33,122 | +316% | 1 | 1 | 0% | 1,592 | 2,948 | +85% | 0 | 0 | — |
case-02 | fail→pass | 6,472 | 3,678 | -43% | 1 | 1 | 0% | 1,070 | 2,974 | +178% | 0 | 0 | — |
case-03 | fail→pass | 6,050 | 3,343 | -45% | 1 | 1 | 0% | 1,082 | 2,950 | +173% | 0 | 0 | — |
case-04 | fail→pass | 42,661 | 1,886 | -96% | 1 | 1 | 0% | 1,182 | 2,626 | +122% | 0 | 0 | — |
case-05 | fail→pass | 8,926 | 2,671 | -70% | 1 | 1 | 0% | 1,483 | 2,786 | +88% | 0 | 0 | — |
case-06 | pass→pass | 7,967 | 1,814 | -77% | 1 | 1 | 0% | 1,280 | 2,646 | +107% | 0 | 0 | — |
case-07 | fail→pass | 10,401 | 2,079 | -80% | 1 | 1 | 0% | 1,689 | 2,633 | +56% | 0 | 0 | — |
case-08 | fail→pass | 7,735 | 1,704 | -78% | 1 | 1 | 0% | 1,231 | 2,575 | +109% | 0 | 0 | — |
case-09 | fail→pass | 9,101 | 2,412 | -73% | 1 | 1 | 0% | 1,365 | 2,794 | +105% | 0 | 0 | — |
case-10 | fail→pass | 7,877 | 2,355 | -70% | 1 | 1 | 0% | 1,141 | 2,697 | +136% | 0 | 0 | — |
case-11 | fail→pass | 7,676 | 1,741 | -77% | 1 | 1 | 0% | 1,096 | 2,588 | +136% | 0 | 0 | — |
case-12 | fail→pass | 3,391 | 3,728 | +10% | 1 | 1 | 0% | 592 | 2,987 | +405% | 0 | 0 | — |
case-13 | fail→pass | 7,317 | 3,029 | -59% | 1 | 1 | 0% | 1,278 | 2,812 | +120% | 0 | 0 | — |
case-14 | fail→pass | 3,625 | 2,504 | -31% | 1 | 1 | 0% | 592 | 2,739 | +363% | 0 | 0 | — |
case-15 | fail→pass | 4,429 | 3,377 | -24% | 1 | 1 | 0% | 694 | 2,732 | +294% | 0 | 0 | — |
case-16 | fail→pass | 6,930 | 2,345 | -66% | 1 | 1 | 0% | 1,232 | 2,696 | +119% | 0 | 0 | — |
case-17 | fail→pass | 5,129 | 2,689 | -48% | 1 | 1 | 0% | 939 | 2,727 | +190% | 0 | 0 | — |
case-18 | pass→pass | 4,934 | 3,610 | -27% | 1 | 1 | 0% | 876 | 2,998 | +242% | 0 | 0 | — |
case-19 | fail→pass | 8,554 | 1,673 | -80% | 1 | 1 | 0% | 1,331 | 2,552 | +92% | 0 | 0 | — |
case-20 | fail→pass | 6,849 | 2,846 | -58% | 1 | 1 | 0% | 1,054 | 2,608 | +147% | 0 | 0 | — |
case-21 | pass→pass | 2,577 | 2,752 | +7% | 1 | 1 | 0% | 415 | 2,644 | +537% | 0 | 0 | — |
case-22 | pass→pass | 3,690 | 2,951 | -20% | 1 | 1 | 0% | 578 | 2,732 | +373% | 0 | 0 | — |
case-23 | pass→pass | 3,009 | 3,138 | +4% | 1 | 1 | 0% | 517 | 2,852 | +452% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted. The headline lift of +78 percentage points is the difference between those two pass rates over the 23 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/17/2026 | +64% |
| gemini-3.6-flash | verified | 8/7/2026 | +59% |
Other measured skills in the registry, with their headline benchmark lift.