Skip to content

[Fix] GitHub mentions fail with LiteLLM helper models #964

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
mrubens merged 2 commits into from
Aug 1, 2026
Merged

[Fix] GitHub mentions fail with LiteLLM helper models #964

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
31a6014
fix: configure LiteLLM for non-task helpers
mrubens Aug 1, 2026
9e614e8
refactor: keep helper model resolution upstream
mrubens Aug 1, 2026
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
245 changes: 2 additions & 243 deletions apps/worker/src/run-task/agent-home.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,6 @@ import {
BEDROCK_MANTLE_OPENAI_OPENCODE_PROVIDER_ID,
BEDROCK_MANTLE_OPENCODE_PROVIDER_ID,
buildInferenceGatewayOpenCodeBaseUrl,
buildOpenAiCompatibleProviderInstance,
buildOpenCodeModelReasoningOptions,
CHATGPT_FAST_MODE_ENV_VAR_NAME,
CHATGPT_GATEWAY_PROVIDER_ID,
Expand All @@ -17,7 +16,7 @@ import {
getInferenceGatewayProvider,
getInferenceGatewayProviderByEnvVarName,
getMcpIntegration,
getOpenAiCompatibleProviderInstance,
getOpenAiCompatibleRuntimeConfigs,
INFERENCE_GATEWAY_CHATGPT_ENV_VAR_NAME,
INFERENCE_GATEWAY_GITHUB_COPILOT_ENV_VAR_NAME,
INFERENCE_GATEWAY_KEYS_ENV_VAR_NAME,
Expand All @@ -27,15 +26,12 @@ import {
XAI_OPENCODE_PROVIDER_ID,
type InferenceGatewayProvider,
isConfiguredEnvValue,
isOpenAiCompatibleProviderId,
isTaskModelIdDisabled,
listOpenAiCompatibleProviderInstancesFromEnvNames,
mergeOpenAiCompatibleProviderConfig,
mergeOpenCodeModelReasoningOptions,
mergeOpenCodeChatGptFastModeOptions,
mergeOpenRouterVariantAliasModels,
normalizeOptionalReasoningEffort,
OPENAI_COMPATIBLE_PROVIDER_ID,
type OpenAiCompatibleProviderInstance,
parseInferenceGatewayKeys,
renderManualSkillMarkdown,
resolveOpenRouterVariantModelAlias,
Expand Down Expand Up @@ -87,152 +83,6 @@ const AZURE_COGNITIVE_SERVICES_PROVIDER_ID = 'azure-cognitive-services';
const AZURE_COGNITIVE_SERVICES_API_KEY_ENV_VAR_NAME =
'AZURE_COGNITIVE_SERVICES_API_KEY';

/** Fallback direct-mode base URL for default and named OpenAI-compatible ids. */
const OPENAI_COMPATIBLE_DEFAULT_FALLBACK_BASE_URL = 'http://127.0.0.1:4000/v1';

const DEFAULT_OPENAI_COMPATIBLE_INSTANCE =
buildOpenAiCompatibleProviderInstance(null);

/**
* Static OpenAI-compatible-style endpoint providers. Catalog entry names and
* env vars for the built-in `openai-compatible` id come from `@roomote/types`;
* ollama/vllm/litellm stay local here. Named `openai-compatible-<slug>`
* connections are discovered dynamically via the shared helpers.
*/
const STATIC_OPENAI_COMPATIBLE_PROVIDER_CONFIGS = {
[OPENAI_COMPATIBLE_PROVIDER_ID]: {
name: DEFAULT_OPENAI_COMPATIBLE_INSTANCE.label,
baseUrlEnvVarName: DEFAULT_OPENAI_COMPATIBLE_INSTANCE.baseUrlEnvVarName,
fallbackBaseUrl: OPENAI_COMPATIBLE_DEFAULT_FALLBACK_BASE_URL,
apiKeyEnvVarName: DEFAULT_OPENAI_COMPATIBLE_INSTANCE.apiKeyEnvVarName as
| string
| undefined,
keyless: false,
// Arbitrary endpoints must not inherit OPENAI_* credentials. When the
// optional dedicated key is blank, omit the API key entirely so keyless
// servers that reject bearer auth match setup qualification behavior.
allowOpenAiEnvFallback: false,
},
ollama: {
name: 'Ollama',
baseUrlEnvVarName: 'OLLAMA_BASE_URL',
fallbackBaseUrl: 'http://127.0.0.1:11434/v1',
apiKeyEnvVarName: undefined as string | undefined,
keyless: true,
allowOpenAiEnvFallback: false,
},
vllm: {
name: 'vLLM',
baseUrlEnvVarName: 'VLLM_BASE_URL',
fallbackBaseUrl: 'http://127.0.0.1:8000/v1',
apiKeyEnvVarName: 'VLLM_API_KEY' as string | undefined,
keyless: false,
allowOpenAiEnvFallback: true,
},
litellm: {
name: 'LiteLLM',
baseUrlEnvVarName: 'LITELLM_BASE_URL',
fallbackBaseUrl: 'http://127.0.0.1:4000/v1',
apiKeyEnvVarName: 'LITELLM_API_KEY' as string | undefined,
keyless: false,
allowOpenAiEnvFallback: true,
},
} as const;

type StaticOpenAiCompatibleProviderId =
keyof typeof STATIC_OPENAI_COMPATIBLE_PROVIDER_CONFIGS;

type OpenAiCompatibleProviderRuntimeConfig = {
name: string;
baseUrlEnvVarName: string;
fallbackBaseUrl: string;
apiKeyEnvVarName: string | undefined;
keyless: boolean;
allowOpenAiEnvFallback: boolean;
};

function toNamedOpenAiCompatibleRuntimeConfig(
instance: OpenAiCompatibleProviderInstance,
): OpenAiCompatibleProviderRuntimeConfig {
return {
name: instance.label,
baseUrlEnvVarName: instance.baseUrlEnvVarName,
fallbackBaseUrl: OPENAI_COMPATIBLE_DEFAULT_FALLBACK_BASE_URL,
apiKeyEnvVarName: instance.apiKeyEnvVarName,
keyless: false,
allowOpenAiEnvFallback: false,
};
}

function resolveNamedOpenAiCompatibleInstance(
providerId: string,
runtimeEnv: Record<string, string>,
): OpenAiCompatibleProviderInstance | null {
const instance = getOpenAiCompatibleProviderInstance(providerId);
if (!instance?.slug) {
return null;
}

if (!instance.labelEnvVarName) {
return instance;
}

const label = runtimeEnv[instance.labelEnvVarName]?.trim();
if (!label) {
return instance;
}

return getOpenAiCompatibleProviderInstance(providerId, { label }) ?? instance;
}

function getOpenAiCompatibleRuntimeConfigs(
modelIds: Array<string | undefined>,
runtimeEnv: Record<string, string>,
): Map<string, OpenAiCompatibleProviderRuntimeConfig> {
const configs = new Map<string, OpenAiCompatibleProviderRuntimeConfig>();

for (const [providerId, provider] of Object.entries(
STATIC_OPENAI_COMPATIBLE_PROVIDER_CONFIGS,
) as Array<
[StaticOpenAiCompatibleProviderId, OpenAiCompatibleProviderRuntimeConfig]
>) {
configs.set(providerId, provider);
}

const candidateProviderIds = new Set<string>();

for (const modelId of modelIds) {
const providerId = modelId?.trim().split('/')[0];
if (providerId && isOpenAiCompatibleProviderId(providerId)) {
candidateProviderIds.add(providerId);
}
}

for (const instance of listOpenAiCompatibleProviderInstancesFromEnvNames(
Object.keys(runtimeEnv),
)) {
candidateProviderIds.add(instance.id);
}

for (const providerId of candidateProviderIds) {
if (configs.has(providerId)) {
continue;
}

const instance = resolveNamedOpenAiCompatibleInstance(
providerId,
runtimeEnv,
);
if (!instance) {
continue;
}

configs.set(providerId, toNamedOpenAiCompatibleRuntimeConfig(instance));
}

return configs;
}

/**
* OpenRouter identifies the calling application through the `HTTP-Referer`
* and `X-Title` request headers rather than the standard `User-Agent`.
Expand Down Expand Up @@ -935,97 +785,6 @@ function toBedrockMantleRuntimeModelId(modelId: string): string {
: modelId;
}

function mergeOpenAiCompatibleProviderConfig(
providerConfig: Record<string, unknown>,
runtimeEnv: Record<string, string>,
modelIds: Array<string | undefined>,
visionModel?: string,
): Record<string, unknown> {
let merged = providerConfig;
const runtimeConfigs = getOpenAiCompatibleRuntimeConfigs(
modelIds,
runtimeEnv,
);

for (const [providerId, provider] of runtimeConfigs) {
const prefix = `${providerId}/`;
const modelIdsForProvider = [
...new Set(
modelIds.flatMap((modelId) => {
const normalized = modelId?.trim();

return normalized?.startsWith(prefix)
? [normalized.slice(prefix.length)]
: [];
}),
),
];

if (modelIdsForProvider.length === 0) {
continue;
}

const existingProvider = asRecord(merged[providerId]);
const existingOptions = asRecord(existingProvider.options);
const existingModels = asRecord(existingProvider.models);
const directApiKey = provider.apiKeyEnvVarName
? runtimeEnv[provider.apiKeyEnvVarName]?.trim()
: undefined;
const baseURL =
runtimeEnv[provider.baseUrlEnvVarName]?.trim() ||
(provider.allowOpenAiEnvFallback
? runtimeEnv.OPENAI_BASE_URL?.trim()
: '') ||
provider.fallbackBaseUrl;
const apiKeyOptions = directApiKey
? { apiKey: `{env:${provider.apiKeyEnvVarName}}` }
: provider.keyless
? { apiKey: 'ollama' }
: provider.allowOpenAiEnvFallback && runtimeEnv.OPENAI_API_KEY?.trim()
? { apiKey: '{env:OPENAI_API_KEY}' }
: {};

merged = {
...merged,
[providerId]: {
...existingProvider,
npm: '@ai-sdk/openai-compatible',
name: provider.name,
options: {
...existingOptions,
baseURL,
// OpenAI-compatible clients require an API key even though Ollama
// itself accepts unauthenticated requests.
...apiKeyOptions,
},
models: {
...existingModels,
...Object.fromEntries(
modelIdsForProvider.map((modelId) => [
modelId,
{
name: modelId,
...(visionModel === `${providerId}/${modelId}`
? {
attachment: true,
modalities: {
input: ['text', 'image'],
output: ['text'],
},
}
: {}),
...asRecord(existingModels[modelId]),
},
]),
),
},
},
};
}

return merged;
}

/**
* When the dequeue env carries an inference gateway URL, rebase each
* gateway-covered provider that a selected model uses onto the gateway. The
Expand Down
45 changes: 45 additions & 0 deletions packages/cloud-agents/src/server/__tests__/opencode-runtime.test.ts
View file
Open in desktop

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

8 changes: 5 additions & 3 deletions packages/cloud-agents/src/server/opencode-runtime.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,6 +7,7 @@ import {
CHATGPT_FAST_MODE_ENV_VAR_NAME,
DISABLED_MODEL_PROVIDER_ENV_VAR_NAMES,
isTaskModelIdDisabled,
mergeOpenAiCompatibleProviderConfig,
mergeOpenCodeModelReasoningOptions,
mergeOpenCodeChatGptFastModeOptions,
mergeOpenRouterVariantAliasModels,
Expand Down Expand Up @@ -86,9 +87,10 @@ function buildModelBackedOpenCodeConfigContent(
smallModel,
])
: providerReasoningConfig;
const providerConfig = mergeOpenRouterVariantAliasModels(
providerModelConfig,
variantAliases,
const providerConfig = mergeOpenAiCompatibleProviderConfig(
mergeOpenRouterVariantAliasModels(providerModelConfig, variantAliases),
env,
[model, smallModel],
);

return JSON.stringify({
Expand Down
Loading
Loading