getcompanion-ai/co-mono
1
0
Fork 0
mirror of https://github.com/getcompanion-ai/co-mono.git synced 2026-04-17 06:04:51 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Merge pull request #344 from default-anton/feature/zai-openai-migration

Browse source 
Migrate zai provider from Anthropic to OpenAI-compatible API
This commit is contained in:
Mario Zechner 2025-12-29 13:37:43 +01:00 • committed by GitHub
commit 557af13042
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
4 changed files with 50 additions and 41 deletions
Download patch file Download diff file Expand all files Collapse all files

48
packages/ai/scripts/generate-models.ts
View file

@ -260,28 +260,34 @@ async function loadModelsDevData(): Promise<Model<any>[]> {
// Process xAi models // Process xAi models
if (data.zai?.models) { if (data.zai?.models) {
for (const [modelId, model] of Object.entries(data.zai.models)) { for (const [modelId, model] of Object.entries(data.zai.models)) {
const m = model as ModelsDevModel; const m = model as ModelsDevModel;
if (m.tool_call !== true) continue; if (m.tool_call !== true) continue;
const supportsImage = m.modalities?.input?.includes("image")
models.push({ models.push({
id: modelId, id: modelId,
name: m.name || modelId, name: m.name || modelId,
api: "anthropic-messages", api: supportsImage ? "openai-completions" : "anthropic-messages",
provider: "zai", provider: "zai",
baseUrl: "https://api.z.ai/api/anthropic", baseUrl: supportsImage ? "https://api.z.ai/api/coding/paas/v4" : "https://api.z.ai/api/anthropic",
reasoning: m.reasoning === true, reasoning: m.reasoning === true,
input: m.modalities?.input?.includes("image") ? ["text", "image"] : ["text"], input: supportsImage ? ["text", "image"] : ["text"],
cost: { cost: {
input: m.cost?.input || 0, input: m.cost?.input || 0,
output: m.cost?.output || 0, output: m.cost?.output || 0,
cacheRead: m.cost?.cache_read || 0, cacheRead: m.cost?.cache_read || 0,
cacheWrite: m.cost?.cache_write || 0, cacheWrite: m.cost?.cache_write || 0,
}, },
contextWindow: m.limit?.context || 4096, ...(supportsImage ? {
maxTokens: m.limit?.output || 4096, compat: {
}); supportsDeveloperRole: false,
} },
} : {}),
contextWindow: m.limit?.context || 4096,
maxTokens: m.limit?.output || 4096,
});
}
} }
// Process Mistral models // Process Mistral models

14
packages/ai/src/models.generated.ts
View file

@ -7029,9 +7029,10 @@ export const MODELS = {
"glm-4.5v": { "glm-4.5v": {
id: "glm-4.5v", id: "glm-4.5v",
name: "GLM-4.5V", name: "GLM-4.5V",
api: "anthropic-messages", api: "openai-completions",
provider: "zai", provider: "zai",
baseUrl: "https://api.z.ai/api/anthropic", baseUrl: "https://api.z.ai/api/coding/paas/v4",
compat: {"supportsDeveloperRole":false},
reasoning: true, reasoning: true,
input: ["text", "image"], input: ["text", "image"],
cost: { cost: {
@ -7042,7 +7043,7 @@ export const MODELS = {
}, },
contextWindow: 64000, contextWindow: 64000,
maxTokens: 16384, maxTokens: 16384,
} satisfies Model<"anthropic-messages">, } satisfies Model<"openai-completions">,
"glm-4.6": { "glm-4.6": {
id: "glm-4.6", id: "glm-4.6",
name: "GLM-4.6", name: "GLM-4.6",
@ -7063,9 +7064,10 @@ export const MODELS = {
"glm-4.6v": { "glm-4.6v": {
id: "glm-4.6v", id: "glm-4.6v",
name: "GLM-4.6V", name: "GLM-4.6V",
api: "anthropic-messages", api: "openai-completions",
provider: "zai", provider: "zai",
baseUrl: "https://api.z.ai/api/anthropic", baseUrl: "https://api.z.ai/api/coding/paas/v4",
compat: {"supportsDeveloperRole":false},
reasoning: true, reasoning: true,
input: ["text", "image"], input: ["text", "image"],
cost: { cost: {
@ -7076,7 +7078,7 @@ export const MODELS = {
}, },
contextWindow: 128000, contextWindow: 128000,
maxTokens: 32768, maxTokens: 32768,
} satisfies Model<"anthropic-messages">, } satisfies Model<"openai-completions">,
"glm-4.7": { "glm-4.7": {
id: "glm-4.7", id: "glm-4.7",
name: "GLM-4.7", name: "GLM-4.7",

12
packages/ai/src/providers/openai-completions.ts
View file

@ -474,10 +474,14 @@ function convertMessages(
// Handle thinking blocks // Handle thinking blocks
const thinkingBlocks = msg.content.filter((b) => b.type === "thinking") as ThinkingContent[]; const thinkingBlocks = msg.content.filter((b) => b.type === "thinking") as ThinkingContent[];
if (thinkingBlocks.length > 0) { // Filter out empty thinking blocks to avoid API validation errors
const nonEmptyThinkingBlocks = thinkingBlocks.filter((b) => b.thinking && b.thinking.trim().length > 0);
if (nonEmptyThinkingBlocks.length > 0) {
if (compat.requiresThinkingAsText) { if (compat.requiresThinkingAsText) {
// Convert thinking blocks to text with <thinking> delimiters // Convert thinking blocks to text with <thinking> delimiters
const thinkingText = thinkingBlocks.map((b) => `<thinking>\n${b.thinking}\n</thinking>`).join("\n"); const thinkingText = nonEmptyThinkingBlocks
.map((b) => `<thinking>\n${b.thinking}\n</thinking>`)
.join("\n");
const textContent = assistantMsg.content as Array<{ type: "text"; text: string }> | null; const textContent = assistantMsg.content as Array<{ type: "text"; text: string }> | null;
if (textContent) { if (textContent) {
textContent.unshift({ type: "text", text: thinkingText }); textContent.unshift({ type: "text", text: thinkingText });
@ -486,9 +490,9 @@ function convertMessages(
} }
} else { } else {
// Use the signature from the first thinking block if available (for llama.cpp server + gpt-oss) // Use the signature from the first thinking block if available (for llama.cpp server + gpt-oss)
const signature = thinkingBlocks[0].thinkingSignature; const signature = nonEmptyThinkingBlocks[0].thinkingSignature;
if (signature && signature.length > 0) { if (signature && signature.length > 0) {
(assistantMsg as any)[signature] = thinkingBlocks.map((b) => b.thinking).join("\n"); (assistantMsg as any)[signature] = nonEmptyThinkingBlocks.map((b) => b.thinking).join("\n");
} }
} }
} }

17
packages/ai/test/stream.test.ts
View file

@ -571,9 +571,8 @@ describe("Generate E2E Tests", () => {
await handleStreaming(llm); await handleStreaming(llm);
}); });
it("should handle thinking", { retry: 3 }, async () => { it.skip("should handle thinking mode", { retry: 3 }, async () => {
// Prompt doesn't trigger thinking await handleThinking(llm, { thinkingEnabled: true, thinkingBudgetTokens: 2048 });
// await handleThinking(llm, { thinkingEnabled: true, thinkingBudgetTokens: 2048 });
}); });
it("should handle multi-turn with thinking and tools", { retry: 3 }, async () => { it("should handle multi-turn with thinking and tools", { retry: 3 }, async () => {
@ -581,7 +580,7 @@ describe("Generate E2E Tests", () => {
}); });
}); });
describe.skipIf(!process.env.ZAI_API_KEY)("zAI Provider (glm-4.5v via Anthropic Messages)", () => { describe.skipIf(!process.env.ZAI_API_KEY)("zAI Provider (glm-4.5v via OpenAI Completions)", () => {
const llm = getModel("zai", "glm-4.5v"); const llm = getModel("zai", "glm-4.5v");
it("should complete basic text generation", { retry: 3 }, async () => { it("should complete basic text generation", { retry: 3 }, async () => {
@ -596,18 +595,16 @@ describe("Generate E2E Tests", () => {
await handleStreaming(llm); await handleStreaming(llm);
}); });
it("should handle thinking", { retry: 3 }, async () => { it("should handle thinking mode", { retry: 3 }, async () => {
// Prompt doesn't trigger thinking await handleThinking(llm, { reasoningEffort: "medium" });
// await handleThinking(llm, { thinkingEnabled: true, thinkingBudgetTokens: 2048 });
}); });
it("should handle multi-turn with thinking and tools", { retry: 3 }, async () => { it("should handle multi-turn with thinking and tools", { retry: 3 }, async () => {
await multiTurn(llm, { thinkingEnabled: true, thinkingBudgetTokens: 2048 }); await multiTurn(llm, { reasoningEffort: "medium" });
}); });
it("should handle image input", { retry: 3 }, async () => { it("should handle image input", { retry: 3 }, async () => {
// Can't see image for some reason? await handleImage(llm);
// await handleImage(llm);
}); });
}); });