feat(model): wire anthropic temperature capability - #1479
Conversation
📝 WalkthroughWalkthroughImplemented temperature control capability detection across the system by adding query methods to Changes
Sequence DiagramsequenceDiagram
participant UI as UI Component
participant ConfigPresenter as ConfigPresenter
participant ModelCapabilities as ModelCapabilities
participant Runtime as AI SDK Runtime
UI->>ConfigPresenter: supportsTemperatureControl(providerId, modelId)
ConfigPresenter->>ModelCapabilities: supportsTemperatureControl(providerId, modelId)
ModelCapabilities->>ModelCapabilities: Check if model matches<br/>unsupported pattern
ModelCapabilities-->>ConfigPresenter: boolean (capability support)
ConfigPresenter-->>UI: boolean (display control?)
alt Temperature Supported
UI->>Runtime: Call with temperature parameter
Runtime->>ConfigPresenter: Verify capability at runtime
ConfigPresenter-->>Runtime: Confirmed true
Runtime->>Runtime: Include temperature in request
else Temperature Not Supported
UI->>Runtime: Call without temperature
Runtime->>ConfigPresenter: Verify capability at runtime
ConfigPresenter-->>Runtime: false or undefined
Runtime->>Runtime: Omit temperature from request
end
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Possibly related PRs
Suggested reviewers
Poem
🚥 Pre-merge checks | ✅ 2 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (2 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
There was a problem hiding this comment.
Actionable comments posted: 5
🧹 Nitpick comments (2)
test/renderer/components/ModelConfigDialog.test.ts (1)
71-73: Fallback path not exercised by this mock shape.
supportsTemperatureControlalways resolves to a boolean (options.temperatureCapability ?? true), so the component code only ever takes the first branch of thetypeof === 'boolean'check. The fallback togetTemperatureCapabilityis never verified. If you want to keep the fallback path meaningful, consider having one mock returnundefined/non-boolean and asserting the other is consulted — otherwise thegetTemperatureCapabilitymock is effectively dead code in every test.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@test/renderer/components/ModelConfigDialog.test.ts` around lines 71 - 73, The mock for supportsTemperatureControl always returns a boolean, so the component's typeof === 'boolean' branch always runs and the fallback to getTemperatureCapability is never exercised; change the mock shape in ModelConfigDialog tests so one scenario returns a non-boolean/undefined from supportsTemperatureControl (via vi.fn().mockResolvedValue(undefined) or similar) and assert that getTemperatureCapability (mocked by getTemperatureCapability) is then called/used; alternatively add an explicit test case that stubs supportsTemperatureControl to a non-boolean and verifies getTemperatureCapability is consulted.src/main/presenter/configPresenter/modelCapabilities.ts (1)
509-512:getTemperatureCapabilityonly looks up the exact provider/model pair.Unlike
getReasoningPortraitwhich falls back across providers viagetModel/registry lookups,getTemperatureCapabilityusesgetProviderMatchonly. When a model is defined in the DB under a canonical provider but queried via an alias (or via a provider that hosts a proxied model without its own DB entry), the capability will beundefinedand the code will rely on the name-pattern fallback. If that's the intended behavior (avoid mixing temperature rules across vendors), it's worth a short comment; otherwise consider delegating togetModellike the reasoning lookups.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/main/presenter/configPresenter/modelCapabilities.ts` around lines 509 - 512, getTemperatureCapability currently calls getProviderMatch(providerId, modelId) and only returns a temperature when there's an exact provider/model DB entry, which misses canonical-provider or alias lookups; either update getTemperatureCapability to delegate to getModel(providerId, modelId) (the same lookup used by getReasoningPortrait) so aliases and registry fallbacks are honored, or if exact-provider semantics are intentional add a concise comment above getTemperatureCapability explaining that it deliberately avoids cross-provider/alias fallbacks; reference getTemperatureCapability, getProviderMatch, getModel, and getReasoningPortrait when making the change.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@resources/model-db/providers.json`:
- Around line 192720-192763: Add the missing "temperature": false flag to the
Claude Opus 4.7 provider entry (identify by id "claude-opus-4-7" / display_name
"claude-opus-4-7") so it matches other Anthropic Claude Opus 4.7 variants;
update the JSON object for that provider to include "temperature": false at the
same top-level position used by other Anthropic entries to ensure the runtime
capability check treats temperature as unsupported.
- Around line 186475-186497: The entry for id
"meta-llama/llama-guard-4-12b:free" is incorrectly marked as "type":
"imageGeneration"; change its "type" to a text/guard/classification type (e.g.,
"text" or "classification") so it is routed to the text/guard pipeline in the
UI, leaving modalities, tool_call, and reasoning unchanged; update only the
"type" field for the object where id equals "meta-llama/llama-guard-4-12b:free".
- Around line 183945-183988: The provider entry for anthropic/claude-opus-4.7 is
incorrectly labeled as an image generation model and missing the temperature
flag; update the object with id "anthropic/claude-opus-4.7" to use the correct
type for chat/reasoning models (change "type": "imageGeneration" to the
chat/text-completion type used elsewhere in the catalog) and add "temperature":
false to the top-level properties so temperature controls are disabled; ensure
reasoning-related fields remain unchanged and that the entry matches the
structure of other chat models in providers.json.
- Around line 164355-164444: The two AIHubMix provider entries with id
"claude-opus-4-7" and "claude-opus-4-7-think" are missing the temperature
capability flag; add "temperature": false to each provider object (same place
other Claude Opus 4.7 entries set it) so the runtime won’t send a temperature
field for these Anthropic-backed models; locate the objects by their id fields
and insert the flag alongside the other top-level provider properties.
In `@src/main/presenter/llmProviderPresenter/aiSdk/runtime.ts`:
- Around line 41-50: The hasAnthropicTemperatureFallback function currently
inspects only modelId and a narrow regex; change its signature to accept the
context (e.g., ModelCapabilities or a small object containing
capabilityProviderId) and early-return false unless capabilityProviderId ===
'anthropic', then broaden ANTHROPIC_TEMPERATURE_UNSUPPORTED_MODEL_PATTERN to
match date-suffixed and optional "-think" variants (suggested pattern:
/^claude-opus-4-7(?:-\d{8})?(?:-think)?$/) and update all call sites to pass the
context so the provider check is enforced; keep the function name and constant
(ANTHROPIC_TEMPERATURE_UNSUPPORTED_MODEL_PATTERN) but update usages accordingly.
---
Nitpick comments:
In `@src/main/presenter/configPresenter/modelCapabilities.ts`:
- Around line 509-512: getTemperatureCapability currently calls
getProviderMatch(providerId, modelId) and only returns a temperature when
there's an exact provider/model DB entry, which misses canonical-provider or
alias lookups; either update getTemperatureCapability to delegate to
getModel(providerId, modelId) (the same lookup used by getReasoningPortrait) so
aliases and registry fallbacks are honored, or if exact-provider semantics are
intentional add a concise comment above getTemperatureCapability explaining that
it deliberately avoids cross-provider/alias fallbacks; reference
getTemperatureCapability, getProviderMatch, getModel, and getReasoningPortrait
when making the change.
In `@test/renderer/components/ModelConfigDialog.test.ts`:
- Around line 71-73: The mock for supportsTemperatureControl always returns a
boolean, so the component's typeof === 'boolean' branch always runs and the
fallback to getTemperatureCapability is never exercised; change the mock shape
in ModelConfigDialog tests so one scenario returns a non-boolean/undefined from
supportsTemperatureControl (via vi.fn().mockResolvedValue(undefined) or similar)
and assert that getTemperatureCapability (mocked by getTemperatureCapability) is
then called/used; alternatively add an explicit test case that stubs
supportsTemperatureControl to a non-boolean and verifies
getTemperatureCapability is consulted.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 298ae968-5098-49b9-b53d-173f33591f11
📒 Files selected for processing (11)
resources/acp-registry/registry.jsonresources/model-db/providers.jsonsrc/main/presenter/configPresenter/index.tssrc/main/presenter/configPresenter/modelCapabilities.tssrc/main/presenter/llmProviderPresenter/aiSdk/runtime.tssrc/renderer/src/components/chat/ChatStatusBar.vuesrc/renderer/src/components/settings/ModelConfigDialog.vuesrc/shared/types/presenters/legacy.presenters.d.tstest/main/presenter/llmProviderPresenter/aiSdkRuntime.test.tstest/renderer/components/ChatStatusBar.test.tstest/renderer/components/ModelConfigDialog.test.ts
| "id": "claude-opus-4-7", | ||
| "name": "claude-opus-4-7", | ||
| "display_name": "claude-opus-4-7", | ||
| "modalities": { | ||
| "input": [ | ||
| "text", | ||
| "image" | ||
| ] | ||
| }, | ||
| "limit": { | ||
| "context": 8192, | ||
| "output": 8192 | ||
| "context": 200000, | ||
| "output": 200000 | ||
| }, | ||
| "tool_call": false, | ||
| "tool_call": true, | ||
| "reasoning": { | ||
| "supported": false | ||
| "supported": true, | ||
| "default": true | ||
| }, | ||
| "extra_capabilities": { | ||
| "reasoning": { | ||
| "supported": true, | ||
| "default_enabled": false, | ||
| "mode": "budget", | ||
| "budget": { | ||
| "min": 1024, | ||
| "unit": "tokens" | ||
| }, | ||
| "interleaved": true, | ||
| "summaries": true, | ||
| "visibility": "summary", | ||
| "continuation": [ | ||
| "thinking_blocks" | ||
| ], | ||
| "notes": [ | ||
| "Anthropic uses thinking budget tokens" | ||
| ] | ||
| } | ||
| }, | ||
| "cost": { | ||
| "input": 2, | ||
| "output": 0, | ||
| "cache_read": 0 | ||
| "input": 5, | ||
| "output": 25, | ||
| "cache_read": 0.5 | ||
| }, | ||
| "type": "imageGeneration" | ||
| "type": "chat" | ||
| }, | ||
| { | ||
| "id": "qwen-image-2.0-pro", | ||
| "name": "qwen-image-2.0-pro", | ||
| "display_name": "qwen-image-2.0-pro", | ||
| "id": "claude-opus-4-7-think", | ||
| "name": "claude-opus-4-7-think", | ||
| "display_name": "claude-opus-4-7-think", | ||
| "modalities": { | ||
| "input": [ | ||
| "text", | ||
| "image" | ||
| "image", | ||
| "text" | ||
| ] | ||
| }, | ||
| "limit": { | ||
| "context": 8192, | ||
| "output": 8192 | ||
| "context": 200000, | ||
| "output": 200000 | ||
| }, | ||
| "tool_call": false, | ||
| "tool_call": true, | ||
| "reasoning": { | ||
| "supported": false | ||
| "supported": true, | ||
| "default": true | ||
| }, | ||
| "extra_capabilities": { | ||
| "reasoning": { | ||
| "supported": true, | ||
| "default_enabled": false, | ||
| "mode": "budget", | ||
| "budget": { | ||
| "min": 1024, | ||
| "unit": "tokens" | ||
| }, | ||
| "interleaved": true, | ||
| "summaries": true, | ||
| "visibility": "summary", | ||
| "continuation": [ | ||
| "thinking_blocks" | ||
| ], | ||
| "notes": [ | ||
| "Anthropic uses thinking budget tokens" | ||
| ] | ||
| } | ||
| }, | ||
| "cost": { | ||
| "input": 2, | ||
| "output": 0, | ||
| "cache_read": 0 | ||
| "input": 5, | ||
| "output": 25, | ||
| "cache_read": 0.5 | ||
| }, | ||
| "type": "imageGeneration" | ||
| "type": "chat" |
There was a problem hiding this comment.
Missing "temperature": false on AIHubMix claude-opus-4-7 and claude-opus-4-7-think.
These are Anthropic Claude Opus 4.7 variants proxied via AIHubMix. All other Claude Opus 4.7 entries added in this PR (e.g., hunks 2, 3, 8, 11, 18-19, 92) correctly set "temperature": false, matching the temperature-capability wiring introduced in modelCapabilities.ts/runtime.ts. These two AIHubMix entries omit the flag, so the runtime will still send temperature for them via AIHubMix, which contradicts the PR's objective and may be rejected by the upstream Anthropic API.
🐛 Proposed fix
{
"id": "claude-opus-4-7",
"name": "claude-opus-4-7",
"display_name": "claude-opus-4-7",
"modalities": {
"input": [
"text",
"image"
]
},
"limit": {
"context": 200000,
"output": 200000
},
+ "temperature": false,
"tool_call": true,
"reasoning": {
...
},
...
"type": "chat"
},
{
"id": "claude-opus-4-7-think",
...
"limit": {
"context": 200000,
"output": 200000
},
+ "temperature": false,
"tool_call": true,📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| "id": "claude-opus-4-7", | |
| "name": "claude-opus-4-7", | |
| "display_name": "claude-opus-4-7", | |
| "modalities": { | |
| "input": [ | |
| "text", | |
| "image" | |
| ] | |
| }, | |
| "limit": { | |
| "context": 8192, | |
| "output": 8192 | |
| "context": 200000, | |
| "output": 200000 | |
| }, | |
| "tool_call": false, | |
| "tool_call": true, | |
| "reasoning": { | |
| "supported": false | |
| "supported": true, | |
| "default": true | |
| }, | |
| "extra_capabilities": { | |
| "reasoning": { | |
| "supported": true, | |
| "default_enabled": false, | |
| "mode": "budget", | |
| "budget": { | |
| "min": 1024, | |
| "unit": "tokens" | |
| }, | |
| "interleaved": true, | |
| "summaries": true, | |
| "visibility": "summary", | |
| "continuation": [ | |
| "thinking_blocks" | |
| ], | |
| "notes": [ | |
| "Anthropic uses thinking budget tokens" | |
| ] | |
| } | |
| }, | |
| "cost": { | |
| "input": 2, | |
| "output": 0, | |
| "cache_read": 0 | |
| "input": 5, | |
| "output": 25, | |
| "cache_read": 0.5 | |
| }, | |
| "type": "imageGeneration" | |
| "type": "chat" | |
| }, | |
| { | |
| "id": "qwen-image-2.0-pro", | |
| "name": "qwen-image-2.0-pro", | |
| "display_name": "qwen-image-2.0-pro", | |
| "id": "claude-opus-4-7-think", | |
| "name": "claude-opus-4-7-think", | |
| "display_name": "claude-opus-4-7-think", | |
| "modalities": { | |
| "input": [ | |
| "text", | |
| "image" | |
| "image", | |
| "text" | |
| ] | |
| }, | |
| "limit": { | |
| "context": 8192, | |
| "output": 8192 | |
| "context": 200000, | |
| "output": 200000 | |
| }, | |
| "tool_call": false, | |
| "tool_call": true, | |
| "reasoning": { | |
| "supported": false | |
| "supported": true, | |
| "default": true | |
| }, | |
| "extra_capabilities": { | |
| "reasoning": { | |
| "supported": true, | |
| "default_enabled": false, | |
| "mode": "budget", | |
| "budget": { | |
| "min": 1024, | |
| "unit": "tokens" | |
| }, | |
| "interleaved": true, | |
| "summaries": true, | |
| "visibility": "summary", | |
| "continuation": [ | |
| "thinking_blocks" | |
| ], | |
| "notes": [ | |
| "Anthropic uses thinking budget tokens" | |
| ] | |
| } | |
| }, | |
| "cost": { | |
| "input": 2, | |
| "output": 0, | |
| "cache_read": 0 | |
| "input": 5, | |
| "output": 25, | |
| "cache_read": 0.5 | |
| }, | |
| "type": "imageGeneration" | |
| "type": "chat" | |
| "id": "claude-opus-4-7", | |
| "name": "claude-opus-4-7", | |
| "display_name": "claude-opus-4-7", | |
| "modalities": { | |
| "input": [ | |
| "text", | |
| "image" | |
| ] | |
| }, | |
| "limit": { | |
| "context": 200000, | |
| "output": 200000 | |
| }, | |
| "temperature": false, | |
| "tool_call": true, | |
| "reasoning": { | |
| "supported": true, | |
| "default": true | |
| }, | |
| "extra_capabilities": { | |
| "reasoning": { | |
| "supported": true, | |
| "default_enabled": false, | |
| "mode": "budget", | |
| "budget": { | |
| "min": 1024, | |
| "unit": "tokens" | |
| }, | |
| "interleaved": true, | |
| "summaries": true, | |
| "visibility": "summary", | |
| "continuation": [ | |
| "thinking_blocks" | |
| ], | |
| "notes": [ | |
| "Anthropic uses thinking budget tokens" | |
| ] | |
| } | |
| }, | |
| "cost": { | |
| "input": 5, | |
| "output": 25, | |
| "cache_read": 0.5 | |
| }, | |
| "type": "chat" | |
| }, | |
| { | |
| "id": "claude-opus-4-7-think", | |
| "name": "claude-opus-4-7-think", | |
| "display_name": "claude-opus-4-7-think", | |
| "modalities": { | |
| "input": [ | |
| "image", | |
| "text" | |
| ] | |
| }, | |
| "limit": { | |
| "context": 200000, | |
| "output": 200000 | |
| }, | |
| "temperature": false, | |
| "tool_call": true, | |
| "reasoning": { | |
| "supported": true, | |
| "default": true | |
| }, | |
| "extra_capabilities": { | |
| "reasoning": { | |
| "supported": true, | |
| "default_enabled": false, | |
| "mode": "budget", | |
| "budget": { | |
| "min": 1024, | |
| "unit": "tokens" | |
| }, | |
| "interleaved": true, | |
| "summaries": true, | |
| "visibility": "summary", | |
| "continuation": [ | |
| "thinking_blocks" | |
| ], | |
| "notes": [ | |
| "Anthropic uses thinking budget tokens" | |
| ] | |
| } | |
| }, | |
| "cost": { | |
| "input": 5, | |
| "output": 25, | |
| "cache_read": 0.5 | |
| }, | |
| "type": "chat" |
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@resources/model-db/providers.json` around lines 164355 - 164444, The two
AIHubMix provider entries with id "claude-opus-4-7" and "claude-opus-4-7-think"
are missing the temperature capability flag; add "temperature": false to each
provider object (same place other Claude Opus 4.7 entries set it) so the runtime
won’t send a temperature field for these Anthropic-backed models; locate the
objects by their id fields and insert the flag alongside the other top-level
provider properties.
| { | ||
| "id": "anthropic/claude-opus-4.7", | ||
| "name": "Anthropic: Claude Opus 4.7", | ||
| "display_name": "Anthropic: Claude Opus 4.7", | ||
| "modalities": { | ||
| "input": [ | ||
| "text", | ||
| "image" | ||
| ], | ||
| "output": [ | ||
| "text" | ||
| ] | ||
| }, | ||
| "limit": { | ||
| "context": 1000000, | ||
| "output": 128000 | ||
| }, | ||
| "tool_call": true, | ||
| "reasoning": { | ||
| "supported": true, | ||
| "default": true | ||
| }, | ||
| "extra_capabilities": { | ||
| "reasoning": { | ||
| "supported": true, | ||
| "default_enabled": false, | ||
| "mode": "budget", | ||
| "budget": { | ||
| "min": 1024, | ||
| "unit": "tokens" | ||
| }, | ||
| "interleaved": true, | ||
| "summaries": true, | ||
| "visibility": "summary", | ||
| "continuation": [ | ||
| "thinking_blocks" | ||
| ], | ||
| "notes": [ | ||
| "Anthropic uses thinking budget tokens" | ||
| ] | ||
| } | ||
| }, | ||
| "type": "imageGeneration" | ||
| }, |
There was a problem hiding this comment.
Wrong type — Claude Opus 4.7 marked as imageGeneration.
anthropic/claude-opus-4.7 is a chat/reasoning model, but this entry sets "type": "imageGeneration". This will cause the model to be mis-routed in the UI and in the providers that consume this catalog (it will appear under image generation rather than chat, and temperature/reasoning wiring for chat paths won't apply). Note this entry also lacks the "temperature": false flag that's central to this PR.
🐛 Proposed fix
"notes": [
"Anthropic uses thinking budget tokens"
]
}
},
- "type": "imageGeneration"
+ "temperature": false,
+ "type": "chat"
},📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| { | |
| "id": "anthropic/claude-opus-4.7", | |
| "name": "Anthropic: Claude Opus 4.7", | |
| "display_name": "Anthropic: Claude Opus 4.7", | |
| "modalities": { | |
| "input": [ | |
| "text", | |
| "image" | |
| ], | |
| "output": [ | |
| "text" | |
| ] | |
| }, | |
| "limit": { | |
| "context": 1000000, | |
| "output": 128000 | |
| }, | |
| "tool_call": true, | |
| "reasoning": { | |
| "supported": true, | |
| "default": true | |
| }, | |
| "extra_capabilities": { | |
| "reasoning": { | |
| "supported": true, | |
| "default_enabled": false, | |
| "mode": "budget", | |
| "budget": { | |
| "min": 1024, | |
| "unit": "tokens" | |
| }, | |
| "interleaved": true, | |
| "summaries": true, | |
| "visibility": "summary", | |
| "continuation": [ | |
| "thinking_blocks" | |
| ], | |
| "notes": [ | |
| "Anthropic uses thinking budget tokens" | |
| ] | |
| } | |
| }, | |
| "type": "imageGeneration" | |
| }, | |
| { | |
| "id": "anthropic/claude-opus-4.7", | |
| "name": "Anthropic: Claude Opus 4.7", | |
| "display_name": "Anthropic: Claude Opus 4.7", | |
| "modalities": { | |
| "input": [ | |
| "text", | |
| "image" | |
| ], | |
| "output": [ | |
| "text" | |
| ] | |
| }, | |
| "limit": { | |
| "context": 1000000, | |
| "output": 128000 | |
| }, | |
| "tool_call": true, | |
| "reasoning": { | |
| "supported": true, | |
| "default": true | |
| }, | |
| "extra_capabilities": { | |
| "reasoning": { | |
| "supported": true, | |
| "default_enabled": false, | |
| "mode": "budget", | |
| "budget": { | |
| "min": 1024, | |
| "unit": "tokens" | |
| }, | |
| "interleaved": true, | |
| "summaries": true, | |
| "visibility": "summary", | |
| "continuation": [ | |
| "thinking_blocks" | |
| ], | |
| "notes": [ | |
| "Anthropic uses thinking budget tokens" | |
| ] | |
| } | |
| }, | |
| "temperature": false, | |
| "type": "chat" | |
| }, |
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@resources/model-db/providers.json` around lines 183945 - 183988, The provider
entry for anthropic/claude-opus-4.7 is incorrectly labeled as an image
generation model and missing the temperature flag; update the object with id
"anthropic/claude-opus-4.7" to use the correct type for chat/reasoning models
(change "type": "imageGeneration" to the chat/text-completion type used
elsewhere in the catalog) and add "temperature": false to the top-level
properties so temperature controls are disabled; ensure reasoning-related fields
remain unchanged and that the entry matches the structure of other chat models
in providers.json.
| { | ||
| "id": "meta-llama/llama-guard-4-12b:free", | ||
| "name": "Meta: Llama Guard 4 12B (free)", | ||
| "display_name": "Meta: Llama Guard 4 12B (free)", | ||
| "modalities": { | ||
| "input": [ | ||
| "image", | ||
| "text" | ||
| ], | ||
| "output": [ | ||
| "text" | ||
| ] | ||
| }, | ||
| "limit": { | ||
| "context": 163840, | ||
| "output": 65000 | ||
| }, | ||
| "tool_call": false, | ||
| "reasoning": { | ||
| "supported": false | ||
| }, | ||
| "type": "imageGeneration" | ||
| }, |
There was a problem hiding this comment.
Wrong type — Llama Guard 4 12B is not an image-generation model.
meta-llama/llama-guard-4-12b:free is a text classification/guard model (text+image input, text output) with tool_call: false and no reasoning. Marking it as "type": "imageGeneration" will misclassify it in the UI and route it through the wrong pipeline.
🐛 Proposed fix
"reasoning": {
"supported": false
},
- "type": "imageGeneration"
+ "type": "chat"
},📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| { | |
| "id": "meta-llama/llama-guard-4-12b:free", | |
| "name": "Meta: Llama Guard 4 12B (free)", | |
| "display_name": "Meta: Llama Guard 4 12B (free)", | |
| "modalities": { | |
| "input": [ | |
| "image", | |
| "text" | |
| ], | |
| "output": [ | |
| "text" | |
| ] | |
| }, | |
| "limit": { | |
| "context": 163840, | |
| "output": 65000 | |
| }, | |
| "tool_call": false, | |
| "reasoning": { | |
| "supported": false | |
| }, | |
| "type": "imageGeneration" | |
| }, | |
| { | |
| "id": "meta-llama/llama-guard-4-12b:free", | |
| "name": "Meta: Llama Guard 4 12B (free)", | |
| "display_name": "Meta: Llama Guard 4 12B (free)", | |
| "modalities": { | |
| "input": [ | |
| "image", | |
| "text" | |
| ], | |
| "output": [ | |
| "text" | |
| ] | |
| }, | |
| "limit": { | |
| "context": 163840, | |
| "output": 65000 | |
| }, | |
| "tool_call": false, | |
| "reasoning": { | |
| "supported": false | |
| }, | |
| "type": "chat" | |
| }, |
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@resources/model-db/providers.json` around lines 186475 - 186497, The entry
for id "meta-llama/llama-guard-4-12b:free" is incorrectly marked as "type":
"imageGeneration"; change its "type" to a text/guard/classification type (e.g.,
"text" or "classification") so it is routed to the text/guard pipeline in the
UI, leaving modalities, tool_call, and reasoning unchanged; update only the
"type" field for the object where id equals "meta-llama/llama-guard-4-12b:free".
| { | ||
| "id": "claude-opus-4-7", | ||
| "name": "claude-opus-4-7", | ||
| "display_name": "claude-opus-4-7", | ||
| "modalities": { | ||
| "input": [ | ||
| "text", | ||
| "image" | ||
| ], | ||
| "output": [ | ||
| "text" | ||
| ] | ||
| }, | ||
| "limit": { | ||
| "context": 1000000, | ||
| "output": 128000 | ||
| }, | ||
| "tool_call": true, | ||
| "reasoning": { | ||
| "supported": true, | ||
| "default": true | ||
| }, | ||
| "extra_capabilities": { | ||
| "reasoning": { | ||
| "supported": true, | ||
| "default_enabled": false, | ||
| "mode": "budget", | ||
| "budget": { | ||
| "min": 1024, | ||
| "unit": "tokens" | ||
| }, | ||
| "interleaved": true, | ||
| "summaries": true, | ||
| "visibility": "summary", | ||
| "continuation": [ | ||
| "thinking_blocks" | ||
| ], | ||
| "notes": [ | ||
| "Anthropic uses thinking budget tokens" | ||
| ] | ||
| } | ||
| }, | ||
| "type": "chat" | ||
| }, |
There was a problem hiding this comment.
Missing "temperature": false on this claude-opus-4-7 entry.
Consistent with the critique on the AIHubMix entries: every Anthropic Claude Opus 4.7 variant in this file sets "temperature": false except this one. If this entry reaches an Anthropic-backed endpoint, the runtime will still include temperature because the capability check reads the flag from the catalog.
🐛 Proposed fix
"limit": {
"context": 1000000,
"output": 128000
},
+ "temperature": false,
"tool_call": true,
"reasoning": {📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| { | |
| "id": "claude-opus-4-7", | |
| "name": "claude-opus-4-7", | |
| "display_name": "claude-opus-4-7", | |
| "modalities": { | |
| "input": [ | |
| "text", | |
| "image" | |
| ], | |
| "output": [ | |
| "text" | |
| ] | |
| }, | |
| "limit": { | |
| "context": 1000000, | |
| "output": 128000 | |
| }, | |
| "tool_call": true, | |
| "reasoning": { | |
| "supported": true, | |
| "default": true | |
| }, | |
| "extra_capabilities": { | |
| "reasoning": { | |
| "supported": true, | |
| "default_enabled": false, | |
| "mode": "budget", | |
| "budget": { | |
| "min": 1024, | |
| "unit": "tokens" | |
| }, | |
| "interleaved": true, | |
| "summaries": true, | |
| "visibility": "summary", | |
| "continuation": [ | |
| "thinking_blocks" | |
| ], | |
| "notes": [ | |
| "Anthropic uses thinking budget tokens" | |
| ] | |
| } | |
| }, | |
| "type": "chat" | |
| }, | |
| { | |
| "id": "claude-opus-4-7", | |
| "name": "claude-opus-4-7", | |
| "display_name": "claude-opus-4-7", | |
| "modalities": { | |
| "input": [ | |
| "text", | |
| "image" | |
| ], | |
| "output": [ | |
| "text" | |
| ] | |
| }, | |
| "limit": { | |
| "context": 1000000, | |
| "output": 128000 | |
| }, | |
| "temperature": false, | |
| "tool_call": true, | |
| "reasoning": { | |
| "supported": true, | |
| "default": true | |
| }, | |
| "extra_capabilities": { | |
| "reasoning": { | |
| "supported": true, | |
| "default_enabled": false, | |
| "mode": "budget", | |
| "budget": { | |
| "min": 1024, | |
| "unit": "tokens" | |
| }, | |
| "interleaved": true, | |
| "summaries": true, | |
| "visibility": "summary", | |
| "continuation": [ | |
| "thinking_blocks" | |
| ], | |
| "notes": [ | |
| "Anthropic uses thinking budget tokens" | |
| ] | |
| } | |
| }, | |
| "type": "chat" | |
| }, |
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@resources/model-db/providers.json` around lines 192720 - 192763, Add the
missing "temperature": false flag to the Claude Opus 4.7 provider entry
(identify by id "claude-opus-4-7" / display_name "claude-opus-4-7") so it
matches other Anthropic Claude Opus 4.7 variants; update the JSON object for
that provider to include "temperature": false at the same top-level position
used by other Anthropic entries to ensure the runtime capability check treats
temperature as unsupported.
| const ANTHROPIC_TEMPERATURE_UNSUPPORTED_MODEL_PATTERN = /^claude-opus-4-7(?:$|-think$)/ | ||
|
|
||
| function hasAnthropicTemperatureFallback(modelId: string): boolean { | ||
| const normalizedModelId = modelId.toLowerCase() | ||
| const unprefixedModelId = normalizedModelId.includes('/') | ||
| ? normalizedModelId.slice(normalizedModelId.lastIndexOf('/') + 1) | ||
| : normalizedModelId | ||
|
|
||
| return ANTHROPIC_TEMPERATURE_UNSUPPORTED_MODEL_PATTERN.test(unprefixedModelId) | ||
| } |
There was a problem hiding this comment.
🧩 Analysis chain
🌐 Web query:
Anthropic Claude Opus 4 model id naming convention date suffix
💡 Result:
Anthropic's Claude model IDs, including for Claude Opus 4, follow a naming convention of "claude-[model_family]-[version]-[YYYYMMDD]" for specific versions, where the date suffix (YYYYMMDD) is a snapshot date indicating the exact model snapshot. This ensures models with the same snapshot date are identical across platforms and do not change, providing stable performance. For example, the initial Claude Opus 4 model ID is "claude-opus-4-20250514" (released May 22, 2025). Later Opus 4 variants include dated suffixes like "claude-opus-4-1-20250805" and "claude-opus-4-5-20251101". Newer models like "claude-opus-4-6" and "claude-opus-4-7" use a simplified "[family]-[version]-[subversion]" format without dates for aliases pointing to the latest, but specific pinned versions retain the date suffix. Aliases like "claude-opus-4-0" resolve to the dated ID "claude-opus-4-20250514".
Citations:
- 1: https://docs.anthropic.com/en/docs/about-claude/models
- 2: https://docs.anthropic.com/en/docs/about-claude/models/all-models
- 3: https://platform.claude.com/docs/en/about-claude/models/overview
- 4: https://docs.anthropic.com/claude/docs/legacy-model-guide
- 5: https://platform.claude.com/docs/en/docs/about-claude/models
🏁 Script executed:
cat -n src/main/presenter/llmProviderPresenter/aiSdk/runtime.ts | head -60Repository: ThinkInAIXYZ/deepchat
Length of output: 2775
🏁 Script executed:
rg 'hasAnthropicTemperatureFallback' --type ts --type tsx -A 5 -B 5Repository: ThinkInAIXYZ/deepchat
Length of output: 92
🏁 Script executed:
rg 'ModelCapabilities' --type ts --type tsx -l | head -5Repository: ThinkInAIXYZ/deepchat
Length of output: 92
🏁 Script executed:
rg 'hasAnthropicTemperatureFallback' -A 5 -B 5Repository: ThinkInAIXYZ/deepchat
Length of output: 1943
🏁 Script executed:
rg 'ModelCapabilities' -B 3 -A 3Repository: ThinkInAIXYZ/deepchat
Length of output: 11761
🏁 Script executed:
sed -n '52,75p' src/main/presenter/llmProviderPresenter/aiSdk/runtime.tsRepository: ThinkInAIXYZ/deepchat
Length of output: 740
🏁 Script executed:
rg 'supportsTemperatureControl' -B 3 -A 3Repository: ThinkInAIXYZ/deepchat
Length of output: 10190
Tighten fallback to Anthropic provider and broaden model pattern.
The hasAnthropicTemperatureFallback function has two fragility issues:
-
Unverified provider: It matches model IDs without confirming the provider is Anthropic. A proxy provider exposing a model id containing
claude-opus-4-7will also strip temperature. Passcontextto the function and gate oncapabilityProviderIdbeing Anthropic. -
Pattern too narrow: The regex
/^claude-opus-4-7(?:$|-think$)/misses date-suffixed variants likeclaude-opus-4-7-YYYYMMDDwhich Anthropic uses in production. Broaden to/^claude-opus-4-7(?:-\d{8})?(?:-think)?$/.
Since ModelCapabilities owns the proper capability logic, consider whether this regex fallback is still needed. If kept, scope it to Anthropic and widen the pattern.
🔧 Suggested fix
-function hasAnthropicTemperatureFallback(modelId: string): boolean {
- const normalizedModelId = modelId.toLowerCase()
- const unprefixedModelId = normalizedModelId.includes('/')
- ? normalizedModelId.slice(normalizedModelId.lastIndexOf('/') + 1)
- : normalizedModelId
-
- return ANTHROPIC_TEMPERATURE_UNSUPPORTED_MODEL_PATTERN.test(unprefixedModelId)
-}
+const ANTHROPIC_TEMPERATURE_UNSUPPORTED_MODEL_PATTERN = /^claude-opus-4-7(?:-\d{8})?(?:-think)?$/
+
+function hasAnthropicTemperatureFallback(
+ context: AiSdkRuntimeContext,
+ modelId: string
+): boolean {
+ const providerId = (context.provider.capabilityProviderId || context.provider.id).toLowerCase()
+ if (!providerId.includes('anthropic')) {
+ return false
+ }
+ const normalizedModelId = modelId.toLowerCase()
+ const unprefixedModelId = normalizedModelId.includes('/')
+ ? normalizedModelId.slice(normalizedModelId.lastIndexOf('/') + 1)
+ : normalizedModelId
+
+ return ANTHROPIC_TEMPERATURE_UNSUPPORTED_MODEL_PATTERN.test(unprefixedModelId)
+}Update the call site to pass context:
- if (hasAnthropicTemperatureFallback(modelId)) {
+ if (hasAnthropicTemperatureFallback(context, modelId)) {📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| const ANTHROPIC_TEMPERATURE_UNSUPPORTED_MODEL_PATTERN = /^claude-opus-4-7(?:$|-think$)/ | |
| function hasAnthropicTemperatureFallback(modelId: string): boolean { | |
| const normalizedModelId = modelId.toLowerCase() | |
| const unprefixedModelId = normalizedModelId.includes('/') | |
| ? normalizedModelId.slice(normalizedModelId.lastIndexOf('/') + 1) | |
| : normalizedModelId | |
| return ANTHROPIC_TEMPERATURE_UNSUPPORTED_MODEL_PATTERN.test(unprefixedModelId) | |
| } | |
| const ANTHROPIC_TEMPERATURE_UNSUPPORTED_MODEL_PATTERN = /^claude-opus-4-7(?:-\d{8})?(?:-think)?$/ | |
| function hasAnthropicTemperatureFallback( | |
| context: AiSdkRuntimeContext, | |
| modelId: string | |
| ): boolean { | |
| const providerId = (context.provider.capabilityProviderId || context.provider.id).toLowerCase() | |
| if (!providerId.includes('anthropic')) { | |
| return false | |
| } | |
| const normalizedModelId = modelId.toLowerCase() | |
| const unprefixedModelId = normalizedModelId.includes('/') | |
| ? normalizedModelId.slice(normalizedModelId.lastIndexOf('/') + 1) | |
| : normalizedModelId | |
| return ANTHROPIC_TEMPERATURE_UNSUPPORTED_MODEL_PATTERN.test(unprefixedModelId) | |
| } |
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@src/main/presenter/llmProviderPresenter/aiSdk/runtime.ts` around lines 41 -
50, The hasAnthropicTemperatureFallback function currently inspects only modelId
and a narrow regex; change its signature to accept the context (e.g.,
ModelCapabilities or a small object containing capabilityProviderId) and
early-return false unless capabilityProviderId === 'anthropic', then broaden
ANTHROPIC_TEMPERATURE_UNSUPPORTED_MODEL_PATTERN to match date-suffixed and
optional "-think" variants (suggested pattern:
/^claude-opus-4-7(?:-\d{8})?(?:-think)?$/) and update all call sites to pass the
context so the provider check is enforced; keep the function name and constant
(ANTHROPIC_TEMPERATURE_UNSUPPORTED_MODEL_PATTERN) but update usages accordingly.
support claude-opus-4-7 and wire anthropic temperature capability
Summary by CodeRabbit
New Features
Chores