102 lines
3.4 KiB
TypeScript
102 lines
3.4 KiB
TypeScript
import { Injectable } from "@nestjs/common";
|
||
import type { LLMProvider } from "./llm-provider.js";
|
||
import type { GenerationContext, GeneratedResponse } from "./types.js";
|
||
import { buildMessages } from "./prompt-builder.js";
|
||
import { extractActions } from "./action-markup.js";
|
||
|
||
interface OpenAiCompatibleChoice {
|
||
message?: { content?: string };
|
||
delta?: { content?: string };
|
||
}
|
||
interface OpenAiCompatibleResponse {
|
||
choices?: OpenAiCompatibleChoice[];
|
||
}
|
||
|
||
function baseUrl(): string {
|
||
return process.env.CLAUDE_BASE_URL ?? "http://localhost:3000/api/v1";
|
||
}
|
||
|
||
function apiKey(): string {
|
||
const key = process.env.CLAUDE_API_KEY;
|
||
if (!key) {
|
||
throw new Error("CLAUDE_API_KEY 未設定，LLM_PROVIDER=claude 需要這個環境變數才能呼叫真實 API");
|
||
}
|
||
return key;
|
||
}
|
||
|
||
function model(): string {
|
||
return process.env.CLAUDE_MODEL ?? "claude-sonnet-4-5";
|
||
}
|
||
|
||
async function requestChatCompletion(body: Record<string, unknown>): Promise<Response> {
|
||
const res = await fetch(`${baseUrl()}/chat/completions`, {
|
||
method: "POST",
|
||
headers: {
|
||
"Content-Type": "application/json",
|
||
Authorization: `Bearer ${apiKey()}`,
|
||
},
|
||
body: JSON.stringify(body),
|
||
});
|
||
if (!res.ok) {
|
||
const text = await res.text().catch(() => "");
|
||
throw new Error(`ClaudeProvider 呼叫失敗（HTTP ${res.status}）：${text.slice(0, 500)}`);
|
||
}
|
||
return res;
|
||
}
|
||
|
||
// R-3：把 F-2 的空殼換成真的呼叫。這個部署走 CLIProxy（分散式派工代理，見
|
||
// gitea.jsc.idv.tw/jiantw83/CLIProxy）提供的 OpenAI 相容端點，而不是直連
|
||
// api.anthropic.com——底層仍是 Claude，只是多一層派工代理，對這個介面
|
||
// （LLMProvider.generate/stream）完全透明，切換供應商不需要動任何呼叫端程式碼。
|
||
@Injectable()
|
||
export class ClaudeProvider implements LLMProvider {
|
||
async generate(context: GenerationContext): Promise<GeneratedResponse> {
|
||
const res = await requestChatCompletion({
|
||
model: model(),
|
||
messages: buildMessages(context),
|
||
stream: false,
|
||
});
|
||
const json = (await res.json()) as OpenAiCompatibleResponse;
|
||
const text = json.choices?.[0]?.message?.content ?? "";
|
||
if (!text) {
|
||
throw new Error("ClaudeProvider 回應內容為空（choices[0].message.content 缺失）");
|
||
}
|
||
return { text, actions: extractActions(text) };
|
||
}
|
||
|
||
async *stream(context: GenerationContext): AsyncIterable<string> {
|
||
const res = await requestChatCompletion({
|
||
model: model(),
|
||
messages: buildMessages(context),
|
||
stream: true,
|
||
});
|
||
if (!res.body) {
|
||
throw new Error("ClaudeProvider 串流回應沒有 body");
|
||
}
|
||
const reader = res.body.getReader();
|
||
const decoder = new TextDecoder();
|
||
let buffer = "";
|
||
while (true) {
|
||
const { done, value } = await reader.read();
|
||
if (done) break;
|
||
buffer += decoder.decode(value, { stream: true });
|
||
const lines = buffer.split("\n");
|
||
buffer = lines.pop() ?? "";
|
||
for (const line of lines) {
|
||
const trimmed = line.trim();
|
||
if (!trimmed.startsWith("data:")) continue;
|
||
const payload = trimmed.slice("data:".length).trim();
|
||
if (payload === "[DONE]") return;
|
||
let chunk: OpenAiCompatibleResponse;
|
||
try {
|
||
chunk = JSON.parse(payload);
|
||
} catch {
|
||
continue;
|
||
}
|
||
const delta = chunk.choices?.[0]?.delta?.content;
|
||
if (delta) yield delta;
|
||
}
|
||
}
|
||
}
|
||
}
|