Skip to content

A self-improving RLM agent for coding workflows and long-running autonomous tasks.

Rust

prefer prime inference for automatic model selection (#422)

ktae44510d909a
diff --git a/packages/coding-agent/CHANGELOG.md b/packages/coding-agent/CHANGELOG.md
index cd58c7f45a..23545c2701 100644
--- a/packages/coding-agent/CHANGELOG.md
+++ b/packages/coding-agent/CHANGELOG.md
@@ -10,6 +10,7 @@
- Changed the shared configuration menu to show prominent, responsive tabs with configurable navigation shortcuts ([ENG-4534](https://linear.app/primeintellect/issue/ENG-4534/make-login-tabs-more-obvious)).
- Fixed Prime Inference login leaving new sessions without a persisted model selection ([ENG-4573](https://linear.app/primeintellect/issue/ENG-4573/prompt-for-model-selection-after-prime-inference-login)).
- Fixed empty prompt placeholders hiding the input caret.
+- Fixed automatic model selection preferring other configured providers over Prime Inference's GLM 5.2 default.
- Fixed missing ripgrep blocking subagents and added actionable installation guidance for the optional search helper ([ENG-4572](https://linear.app/primeintellect/issue/ENG-4572/ripgrep-not-installed)).
- Fixed the resume picker opening on an older session instead of the newest session ([ENG-4630](https://linear.app/primeintellect/issue/ENG-4630/show-latest-sessions-first-in-resume-list)).
- Fixed tool-only responses rendering directly against the preceding user prompt.
diff --git a/packages/coding-agent/src/core/model-resolver.ts b/packages/coding-agent/src/core/model-resolver.ts
index 2ec9f16339..4f06123a75 100644
--- a/packages/coding-agent/src/core/model-resolver.ts
+++ b/packages/coding-agent/src/core/model-resolver.ts
@@ -177,6 +177,25 @@ function buildFallbackModel(provider: string, modelId: string, availableModels:
};
}
+function findPreferredDefaultModel(availableModels: Model<Api>[]): Model<Api> | undefined {
+ const primeInferenceDefault = availableModels.find(
+ (model) => model.provider === "prime-inference" && model.id === PRIME_INFERENCE_DEFAULT_MODEL_ID,
+ );
+ if (primeInferenceDefault) {
+ return primeInferenceDefault;
+ }
+
+ for (const provider of Object.keys(defaultModelPerProvider) as KnownProvider[]) {
+ const defaultId = defaultModelPerProvider[provider];
+ const match = availableModels.find((model) => model.provider === provider && model.id === defaultId);
+ if (match) {
+ return match;
+ }
+ }
+
+ return undefined;
+}
+
/**
* Parse a pattern to extract model and thinking level.
* Handles models with colons in their IDs (e.g., OpenRouter's :exacto suffix).
@@ -561,13 +580,9 @@ export async function findInitialModel(options: {
const availableModels = await modelRegistry.getAvailable();
if (availableModels.length > 0) {
- // Try to find a default model from known providers
- for (const provider of Object.keys(defaultModelPerProvider) as KnownProvider[]) {
- const defaultId = defaultModelPerProvider[provider];
- const match = availableModels.find((m) => m.provider === provider && m.id === defaultId);
- if (match) {
- return { model: match, thinkingLevel: DEFAULT_THINKING_LEVEL, fallbackMessage: undefined };
- }
+ const defaultModel = findPreferredDefaultModel(availableModels);
+ if (defaultModel) {
+ return { model: defaultModel, thinkingLevel: DEFAULT_THINKING_LEVEL, fallbackMessage: undefined };
}
// If no default found, use first available
@@ -623,21 +638,7 @@ export async function restoreModelFromSession(
const availableModels = await modelRegistry.getAvailable();
if (availableModels.length > 0) {
- // Try to find a default model from known providers
- let fallbackModel: Model<Api> | undefined;
- for (const provider of Object.keys(defaultModelPerProvider) as KnownProvider[]) {
- const defaultId = defaultModelPerProvider[provider];
- const match = availableModels.find((m) => m.provider === provider && m.id === defaultId);
- if (match) {
- fallbackModel = match;
- break;
- }
- }
-
- // If no default found, use first available
- if (!fallbackModel) {
- fallbackModel = availableModels[0];
- }
+ const fallbackModel = findPreferredDefaultModel(availableModels) ?? availableModels[0];
if (shouldPrintMessages) {
console.log(chalk.dim(`Falling back to: ${fallbackModel.provider}/${fallbackModel.id}`));
diff --git a/packages/coding-agent/test/model-resolver.test.ts b/packages/coding-agent/test/model-resolver.test.ts
index 3ca1851c5a..511260d430 100644
--- a/packages/coding-agent/test/model-resolver.test.ts
+++ b/packages/coding-agent/test/model-resolver.test.ts
@@ -110,6 +110,26 @@ describe("parseModelPattern", () => {
expect(result.warning).toBeUndefined();
});
+ test("preserves provider-qualified selections when model names overlap", () => {
+ const primeInferenceModel: Model<"anthropic-messages"> = {
+ ...mockModels[0],
+ id: "z-ai/glm-5.2",
+ name: "GLM 5.2",
+ provider: "prime-inference",
+ baseUrl: "https://api.pinference.ai/api/v1",
+ };
+ const huggingFaceModel: Model<"anthropic-messages"> = {
+ ...primeInferenceModel,
+ id: "zai-org/GLM-5.2",
+ provider: "huggingface",
+ baseUrl: "https://router.huggingface.co/v1",
+ };
+
+ const result = parseModelPattern("huggingface/zai-org/GLM-5.2", [primeInferenceModel, huggingFaceModel]);
+
+ expect(result.model).toBe(huggingFaceModel);
+ });
+
test("no match returns undefined model and thinking level", () => {
const result = parseModelPattern("nonexistent", allModels);
expect(result.model).toBeUndefined();
@@ -452,7 +472,12 @@ describe("default model selection", () => {
expect(result.thinkingLevel).toBe("medium");
});
- test("findInitialModel selects GLM 5.2 as the Prime Inference default", async () => {
+ test("findInitialModel prefers GLM 5.2 when Prime Inference is configured", async () => {
+ const anthropicModel: Model<"anthropic-messages"> = {
+ ...mockModels[0],
+ id: "claude-opus-4-7",
+ name: "Claude Opus 4.7",
+ };
const primeModel: Model<"anthropic-messages"> = {
id: "z-ai/glm-5.2",
name: "GLM 5.2",
@@ -466,7 +491,7 @@ describe("default model selection", () => {
maxTokens: 101376,
};
const registry = {
- getAvailable: async () => [primeModel],
+ getAvailable: async () => [anthropicModel, primeModel],
} as unknown as Parameters<typeof findInitialModel>[0]["modelRegistry"];
const result = await findInitialModel({
@@ -478,6 +503,25 @@ describe("default model selection", () => {
expect(result.model).toBe(primeModel);
});
+ test("findInitialModel uses another provider default when Prime Inference is not configured", async () => {
+ const anthropicModel: Model<"anthropic-messages"> = {
+ ...mockModels[0],
+ id: "claude-opus-4-7",
+ name: "Claude Opus 4.7",
+ };
+ const registry = {
+ getAvailable: async () => [anthropicModel],
+ } as unknown as Parameters<typeof findInitialModel>[0]["modelRegistry"];
+
+ const result = await findInitialModel({
+ scopedModels: [],
+ isContinuing: false,
+ modelRegistry: registry,
+ });
+
+ expect(result.model).toBe(anthropicModel);
+ });
+
test("findInitialModel selects ai-gateway default when available", async () => {
const aiGatewayModel: Model<"anthropic-messages"> = {
id: "anthropic/claude-opus-4-6",