Skip to content

feat(presets-parity): assert contextWindows against /model/info; align presets (OPS-732) #43

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
todie merged 1 commit into from
Jul 18, 2026
Merged

feat(presets-parity): assert contextWindows against /model/info; align presets (OPS-732) #43

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
34 changes: 17 additions & 17 deletions packages/harness-omp/config/models.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,15 +17,15 @@ providers:
models:
# crusoe
- id: zai/GLM-5.2
contextWindow: 200000
contextWindow: 1000000
reasoning: true
- id: deepseek-ai/DeepSeek-V4-Pro
contextWindow: 131072
contextWindow: 1000000
reasoning: true
- id: Qwen/Qwen3-235B-A22B-Instruct-2507
contextWindow: 262144
- id: moonshotai/Kimi-K2.6
contextWindow: 131072
contextWindow: 262144
# openrouter (context per live catalog 2026-07-07)
- id: openai/gpt-5.5
contextWindow: 1050000
Expand Down Expand Up @@ -73,43 +73,43 @@ providers:
# mirror their base model's contextWindow; new-model windows are the
# families' published values.
- id: glm-5.2
contextWindow: 200000
contextWindow: 1000000
reasoning: true
- id: glm-5.2-fast
contextWindow: 200000
contextWindow: 1000000
reasoning: true
- id: glm-5.2-flex
contextWindow: 200000
contextWindow: 1000000
reasoning: true
- id: glm-5.2-short
contextWindow: 200000
contextWindow: 199984
reasoning: true
- id: glm-5.2-short-fast
contextWindow: 200000
contextWindow: 199984
reasoning: true
- id: glm-5.2-short-fast-flex
contextWindow: 200000
contextWindow: 199984
reasoning: true
- id: glm-5.2-short-flex
contextWindow: 200000
contextWindow: 199984
reasoning: true
- id: zai/GLM-5.1
contextWindow: 200000
reasoning: true
- id: kimi-k2.6
contextWindow: 131072
contextWindow: 262144
- id: kimi-k2.6-fast
contextWindow: 131072
contextWindow: 262144
- id: kimi-k2.6-flex
contextWindow: 131072
contextWindow: 262144
- id: kimi-k2.7-code
contextWindow: 262144
- id: kimi-k2.7-code-flex
contextWindow: 262144
- id: deepseek-ai/DeepSeek-V3-0324
contextWindow: 131072
contextWindow: 163840
- id: deepseek-ai/Deepseek-V4-Flash
contextWindow: 131072
contextWindow: 1000000
reasoning: true
- id: google/gemma-4-31b-it
contextWindow: 131072
Expand All @@ -126,9 +126,9 @@ providers:
- id: qwen3.5-397b-fast
contextWindow: 262144
- id: qwen3.6-35b
contextWindow: 262144
contextWindow: 131072
- id: qwen3.6-35b-fast
contextWindow: 262144
contextWindow: 131072
- id: yutori/n1.5
contextWindow: 131072
ollama:
Expand Down
263 changes: 215 additions & 48 deletions packages/harness-pi/config/models.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,75 +5,242 @@
"api": "openai-completions",
"apiKey": "env:UNSIGNED_LLM_API_KEY",
"models": [
{ "id": "google/gemini-3.5-flash", "contextWindow": 1048576, "reasoning": true },
{ "id": "zai/GLM-5.2", "contextWindow": 200000, "reasoning": true },
{ "id": "deepseek-ai/DeepSeek-V4-Pro", "contextWindow": 131072, "reasoning": true },
{ "id": "Qwen/Qwen3-235B-A22B-Instruct-2507", "contextWindow": 262144 },
{ "id": "moonshotai/Kimi-K2.6", "contextWindow": 131072 },
{ "id": "openai/gpt-5.5", "contextWindow": 1050000, "reasoning": true },
{ "id": "moonshotai/kimi-k2.7-code", "contextWindow": 262144 },
{ "id": "mistralai/devstral-2512", "contextWindow": 262144 },
{ "id": "claude-sonnet-5", "contextWindow": 200000, "reasoning": true },
{ "id": "claude-haiku-4-5", "contextWindow": 200000 },
{ "id": "claude-opus-4-8", "contextWindow": 200000, "reasoning": true },
{ "id": "morph-v3-fast", "contextWindow": 32768 },
{ "id": "morph-glm52-744b", "contextWindow": 1000000 },
{ "id": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B", "contextWindow": 131072, "reasoning": true },
{ "id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B", "contextWindow": 131072, "reasoning": true },
{ "id": "nvidia/NVIDIA-Nemotron-3-Ultra-550B", "contextWindow": 200000, "reasoning": true },
{ "id": "nvidia/Nemotron-3-Nano-Omni-Reasoning-30B-A3B", "contextWindow": 131072, "reasoning": true },
{ "id": "glm-5.2", "contextWindow": 200000, "reasoning": true },
{ "id": "glm-5.2-fast", "contextWindow": 200000, "reasoning": true },
{ "id": "glm-5.2-flex", "contextWindow": 200000, "reasoning": true },
{ "id": "glm-5.2-short", "contextWindow": 200000, "reasoning": true },
{ "id": "glm-5.2-short-fast", "contextWindow": 200000, "reasoning": true },
{ "id": "glm-5.2-short-fast-flex", "contextWindow": 200000, "reasoning": true },
{ "id": "glm-5.2-short-flex", "contextWindow": 200000, "reasoning": true },
{ "id": "zai/GLM-5.1", "contextWindow": 200000, "reasoning": true },
{ "id": "kimi-k2.6", "contextWindow": 131072 },
{ "id": "kimi-k2.6-fast", "contextWindow": 131072 },
{ "id": "kimi-k2.6-flex", "contextWindow": 131072 },
{ "id": "kimi-k2.7-code", "contextWindow": 262144 },
{ "id": "kimi-k2.7-code-flex", "contextWindow": 262144 },
{ "id": "deepseek-ai/DeepSeek-V3-0324", "contextWindow": 131072 },
{ "id": "deepseek-ai/Deepseek-V4-Flash", "contextWindow": 131072, "reasoning": true },
{ "id": "google/gemma-4-31b-it", "contextWindow": 131072 },
{ "id": "meta-llama/Llama-3.3-70B-Instruct", "contextWindow": 131072 },
{ "id": "minimax/minimax-m2.7", "contextWindow": 200000, "reasoning": true },
{ "id": "openai/gpt-oss-120b", "contextWindow": 131072, "reasoning": true },
{ "id": "qwen3.5-397b", "contextWindow": 262144 },
{ "id": "qwen3.5-397b-fast", "contextWindow": 262144 },
{ "id": "qwen3.6-35b", "contextWindow": 262144 },
{ "id": "qwen3.6-35b-fast", "contextWindow": 262144 },
{ "id": "yutori/n1.5", "contextWindow": 131072 }
{
"id": "google/gemini-3.5-flash",
"contextWindow": 1048576,
"reasoning": true
},
{
"id": "zai/GLM-5.2",
"contextWindow": 1000000,
"reasoning": true
},
{
"id": "deepseek-ai/DeepSeek-V4-Pro",
"contextWindow": 1000000,
"reasoning": true
},
{
"id": "Qwen/Qwen3-235B-A22B-Instruct-2507",
"contextWindow": 262144
},
{
"id": "moonshotai/Kimi-K2.6",
"contextWindow": 262144
},
{
"id": "openai/gpt-5.5",
"contextWindow": 1050000,
"reasoning": true
},
{
"id": "moonshotai/kimi-k2.7-code",
"contextWindow": 262144
},
{
"id": "mistralai/devstral-2512",
"contextWindow": 262144
},
{
"id": "claude-sonnet-5",
"contextWindow": 200000,
"reasoning": true
},
{
"id": "claude-haiku-4-5",
"contextWindow": 200000
},
{
"id": "claude-opus-4-8",
"contextWindow": 200000,
"reasoning": true
},
{
"id": "morph-v3-fast",
"contextWindow": 32768
},
{
"id": "morph-glm52-744b",
"contextWindow": 1000000
},
{
"id": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B",
"contextWindow": 131072,
"reasoning": true
},
{
"id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B",
"contextWindow": 131072,
"reasoning": true
},
{
"id": "nvidia/NVIDIA-Nemotron-3-Ultra-550B",
"contextWindow": 200000,
"reasoning": true
},
{
"id": "nvidia/Nemotron-3-Nano-Omni-Reasoning-30B-A3B",
"contextWindow": 131072,
"reasoning": true
},
{
"id": "glm-5.2",
"contextWindow": 1000000,
"reasoning": true
},
{
"id": "glm-5.2-fast",
"contextWindow": 1000000,
"reasoning": true
},
{
"id": "glm-5.2-flex",
"contextWindow": 1000000,
"reasoning": true
},
{
"id": "glm-5.2-short",
"contextWindow": 199984,
"reasoning": true
},
{
"id": "glm-5.2-short-fast",
"contextWindow": 199984,
"reasoning": true
},
{
"id": "glm-5.2-short-fast-flex",
"contextWindow": 199984,
"reasoning": true
},
{
"id": "glm-5.2-short-flex",
"contextWindow": 199984,
"reasoning": true
},
{
"id": "zai/GLM-5.1",
"contextWindow": 200000,
"reasoning": true
},
{
"id": "kimi-k2.6",
"contextWindow": 262144
},
{
"id": "kimi-k2.6-fast",
"contextWindow": 262144
},
{
"id": "kimi-k2.6-flex",
"contextWindow": 262144
},
{
"id": "kimi-k2.7-code",
"contextWindow": 262144
},
{
"id": "kimi-k2.7-code-flex",
"contextWindow": 262144
},
{
"id": "deepseek-ai/DeepSeek-V3-0324",
"contextWindow": 163840
},
{
"id": "deepseek-ai/Deepseek-V4-Flash",
"contextWindow": 1000000,
"reasoning": true
},
{
"id": "google/gemma-4-31b-it",
"contextWindow": 131072
},
{
"id": "meta-llama/Llama-3.3-70B-Instruct",
"contextWindow": 131072
},
{
"id": "minimax/minimax-m2.7",
"contextWindow": 200000,
"reasoning": true
},
{
"id": "openai/gpt-oss-120b",
"contextWindow": 131072,
"reasoning": true
},
{
"id": "qwen3.5-397b",
"contextWindow": 262144
},
{
"id": "qwen3.5-397b-fast",
"contextWindow": 262144
},
{
"id": "qwen3.6-35b",
"contextWindow": 131072
},
{
"id": "qwen3.6-35b-fast",
"contextWindow": 131072
},
{
"id": "yutori/n1.5",
"contextWindow": 131072
}
]
},
"ollama": {
"baseUrl": "http://localhost:11434/v1",
"api": "openai-completions",
"apiKey": "ollama",
"models": [
{ "id": "qwen3.6-27b-mtp", "contextWindow": 32768, "reasoning": true },
{ "id": "qwen3.6-35b-mtp", "contextWindow": 32768, "reasoning": true },
{ "id": "hf.co/unsloth/GLM-4.7-Flash-GGUF:UD-Q4_K_XL", "contextWindow": 32768, "reasoning": true },
{ "id": "hf.co/unsloth/Devstral-Small-2-24B-Instruct-2512-GGUF:UD-Q4_K_XL", "contextWindow": 32768 },
{ "id": "qwen3.5-9b-mtp", "contextWindow": 32768, "reasoning": true }
{
"id": "qwen3.6-27b-mtp",
"contextWindow": 32768,
"reasoning": true
},
{
"id": "qwen3.6-35b-mtp",
"contextWindow": 32768,
"reasoning": true
},
{
"id": "hf.co/unsloth/GLM-4.7-Flash-GGUF:UD-Q4_K_XL",
"contextWindow": 32768,
"reasoning": true
},
{
"id": "hf.co/unsloth/Devstral-Small-2-24B-Instruct-2512-GGUF:UD-Q4_K_XL",
"contextWindow": 32768
},
{
"id": "qwen3.5-9b-mtp",
"contextWindow": 32768,
"reasoning": true
}
]
},
"llamacpp": {
"baseUrl": "http://localhost:8080/v1",
"api": "openai-completions",
"apiKey": "none",
"models": [
{ "id": "local" }
{
"id": "local"
}
]
},
"vllm": {
"baseUrl": "http://localhost:8000/v1",
"api": "openai-completions",
"apiKey": "none",
"models": [
{ "id": "served-model" }
{
"id": "served-model"
}
]
}
}
Expand Down
11 changes: 6 additions & 5 deletions packages/presets-parity/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,11 +26,12 @@ on a weekly schedule (`.github/workflows/parity-live.yml`, Mon 06:17 UTC +
(gateway virtual key; 1P item `unsigned-llm` — never committed). A red
scheduled run means the gateway catalog drifted from the presets (OPS-729).

The live run also compares preset `contextWindow`s against the gateway's
`/v1/model/info` metadata (OPS-732): mismatches are **warned**, not failed,
while `max_input_tokens` coverage is sparse. Once the LiteLLM config carries
metadata for the full catalog (paas-side half of OPS-732), promote the
warning to an assertion.
The live run also **asserts** preset `contextWindow`s equal the gateway's
`/v1/model/info` `max_input_tokens` (OPS-732; authoritative since
unsigned-paas#1449 / chart 0.4.13 enriched all 41 catalog models), with a
coverage floor so a silent metadata collapse can't make the check vacuous.
On drift: fix the paas values first if the gateway is wrong, else align the
presets — /model/info is the source of truth.

## Verified

Expand Down
Loading