aiwcli 0.12.6 → 0.12.8

package/dist/templates/cc-native/_cc-native/lib-ts/rlm/ollama-client.ts

@@ -1,208 +1,208 @@
-/**
- * Ollama HTTP client for local embeddings.
- *
- * Uses nomic-embed-text via Ollama's /api/embed endpoint.
- * Zero API cost, fast, private.
- */
-
-import { z } from "zod";
-import {
-  OLLAMA_BASE_URL,
-  OLLAMA_EMBED_MODEL,
-  EMBED_DIMENSIONS,
-  HYDE_OLLAMA_MODEL,
-} from "./types.js";
-import { logDebug, logError } from "./logger.js";
-
-const HOOK_NAME = "rlm_ollama";
-const BATCH_SIZE = 32;
-
-// Zod schemas for runtime validation
-const OllamaTagsResponseSchema = z.object({
-  models: z.array(z.object({ name: z.string() })).optional(),
-});
-
-const OllamaEmbedResponseSchema = z.object({
-  embeddings: z.array(z.array(z.number())),
-});
-
-export interface OllamaConfig {
-  baseUrl: string;
-  model: string;
-  dimensions: number;
-}
-
-const DEFAULT_CONFIG: OllamaConfig = {
-  baseUrl: OLLAMA_BASE_URL,
-  model: OLLAMA_EMBED_MODEL,
-  dimensions: EMBED_DIMENSIONS,
-};
-
-export interface HealthResult {
-  ok: boolean;
-  error?: string;
-}
-
-/**
- * Check if Ollama is running and the embedding model is available.
- */
-export async function checkOllamaHealth(
-  config: Partial<OllamaConfig> = {},
-): Promise<HealthResult> {
-  const cfg = { ...DEFAULT_CONFIG, ...config };
-  try {
-    const resp = await fetch(`${cfg.baseUrl}/api/tags`, {
-      signal: AbortSignal.timeout(5_000),
-    });
-    if (!resp.ok) {
-      return { ok: false, error: `Ollama responded with ${resp.status}` };
-    }
-    const json = await resp.json();
-    const parseResult = OllamaTagsResponseSchema.safeParse(json);
-    if (!parseResult.success) {
-      return { ok: false, error: `Invalid Ollama API response: ${parseResult.error.message}` };
-    }
-    const models = parseResult.data.models ?? [];
-    const found = models.some(
-      (m) => m.name === cfg.model || m.name.startsWith(`${cfg.model}:`),
-    );
-    if (!found) {
-      const available = models.map((m) => m.name).join(", ") || "none";
-      return {
-        ok: false,
-        error: `Model "${cfg.model}" not found. Available: ${available}. Run: ollama pull ${cfg.model}`,
-      };
-    }
-    return { ok: true };
-  } catch (e: unknown) {
-    const msg = e instanceof Error ? e.message : String(e);
-    return {
-      ok: false,
-      error: `Cannot reach Ollama at ${cfg.baseUrl}: ${msg}. Is Ollama running?`,
-    };
-  }
-}
-
-/**
- * Embed multiple texts via Ollama /api/embed.
- * Batches at BATCH_SIZE (Ollama processes sequentially internally).
- * Returns one Float32Array per input text.
- */
-export async function embed(
-  texts: string[],
-  config: Partial<OllamaConfig> = {},
-): Promise<Float32Array[]> {
-  const cfg = { ...DEFAULT_CONFIG, ...config };
-  const results: Float32Array[] = [];
-
-  for (let i = 0; i < texts.length; i += BATCH_SIZE) {
-    const batch = texts.slice(i, i + BATCH_SIZE);
-    logDebug(HOOK_NAME, `Embedding batch ${i / BATCH_SIZE + 1} (${batch.length} texts)`);
-
-    const resp = await fetch(`${cfg.baseUrl}/api/embed`, {
-      method: "POST",
-      headers: { "Content-Type": "application/json" },
-      body: JSON.stringify({ model: cfg.model, input: batch }),
-      signal: AbortSignal.timeout(30_000),
-    });
-
-    if (!resp.ok) {
-      const body = await resp.text().catch(() => "");
-      throw new Error(`Ollama embed failed (${resp.status}): ${body}`);
-    }
-
-    const json = await resp.json();
-    const parseResult = OllamaEmbedResponseSchema.safeParse(json);
-    if (!parseResult.success) {
-      throw new Error(`Invalid Ollama embed response: ${parseResult.error.message}`);
-    }
-
-    const data = parseResult.data;
-    if (data.embeddings.length !== batch.length) {
-      throw new Error(
-        `Expected ${batch.length} embeddings, got ${data.embeddings.length}`,
-      );
-    }
-
-    for (const vec of data.embeddings) {
-      results.push(new Float32Array(vec));
-    }
-  }
-
-  return results;
-}
-
-/**
- * Embed a single text. Convenience wrapper around embed().
- */
-export async function embedOne(
-  text: string,
-  config: Partial<OllamaConfig> = {},
-): Promise<Float32Array> {
-  const results = await embed([text], config);
-  if (results.length === 0) {
-    throw new Error("Embedding failed: received empty result array");
-  }
-  return results[0];
-}
-
-/**
- * Generate text using Ollama's /api/generate endpoint.
- * Follows same pattern as embed()/embedOne() for consistency.
- */
-export async function generateText(
-  prompt: string,
-  options?: {
-    systemPrompt?: string;
-    temperature?: number;
-    maxTokens?: number;
-    timeout?: number;
-    model?: string;
-  },
-): Promise<{ success: boolean; text: string; error?: string; latency_ms: number }> {
-  const startTime = Date.now();
-  const model = options?.model ?? HYDE_OLLAMA_MODEL;
-  const baseUrl = process.env.OLLAMA_BASE_URL ?? OLLAMA_BASE_URL;
-
-  try {
-    const response = await fetch(`${baseUrl}/api/generate`, {
-      method: "POST",
-      headers: { "Content-Type": "application/json" },
-      body: JSON.stringify({
-        model,
-        prompt: options?.systemPrompt
-          ? `${options.systemPrompt}\n\n${prompt}`
-          : prompt,
-        stream: false, // Non-streaming for simplicity
-        options: {
-          temperature: options?.temperature ?? 0.7,
-          num_predict: options?.maxTokens ?? 200,
-        },
-      }),
-      signal: AbortSignal.timeout(options?.timeout ?? 10_000),
-    });
-
-    if (!response.ok) {
-      return {
-        success: false,
-        text: "",
-        error: `HTTP ${response.status}: ${response.statusText}`,
-        latency_ms: Date.now() - startTime,
-      };
-    }
-
-    const data = await response.json();
-    return {
-      success: true,
-      text: data.response || "",
-      latency_ms: Date.now() - startTime,
-    };
-  } catch (e) {
-    return {
-      success: false,
-      text: "",
-      error: String(e),
-      latency_ms: Date.now() - startTime,
-    };
-  }
-}
+/**
+ * Ollama HTTP client for local embeddings.
+ *
+ * Uses nomic-embed-text via Ollama's /api/embed endpoint.
+ * Zero API cost, fast, private.
+ */
+
+import { z } from "zod";
+import {
+  OLLAMA_BASE_URL,
+  OLLAMA_EMBED_MODEL,
+  EMBED_DIMENSIONS,
+  HYDE_OLLAMA_MODEL,
+} from "./types.js";
+import { logDebug, logError } from "./logger.js";
+
+const HOOK_NAME = "rlm_ollama";
+const BATCH_SIZE = 32;
+
+// Zod schemas for runtime validation
+const OllamaTagsResponseSchema = z.object({
+  models: z.array(z.object({ name: z.string() })).optional(),
+});
+
+const OllamaEmbedResponseSchema = z.object({
+  embeddings: z.array(z.array(z.number())),
+});
+
+export interface OllamaConfig {
+  baseUrl: string;
+  model: string;
+  dimensions: number;
+}
+
+const DEFAULT_CONFIG: OllamaConfig = {
+  baseUrl: OLLAMA_BASE_URL,
+  model: OLLAMA_EMBED_MODEL,
+  dimensions: EMBED_DIMENSIONS,
+};
+
+export interface HealthResult {
+  ok: boolean;
+  error?: string;
+}
+
+/**
+ * Check if Ollama is running and the embedding model is available.
+ */
+export async function checkOllamaHealth(
+  config: Partial<OllamaConfig> = {},
+): Promise<HealthResult> {
+  const cfg = { ...DEFAULT_CONFIG, ...config };
+  try {
+    const resp = await fetch(`${cfg.baseUrl}/api/tags`, {
+      signal: AbortSignal.timeout(5_000),
+    });
+    if (!resp.ok) {
+      return { ok: false, error: `Ollama responded with ${resp.status}` };
+    }
+    const json = await resp.json();
+    const parseResult = OllamaTagsResponseSchema.safeParse(json);
+    if (!parseResult.success) {
+      return { ok: false, error: `Invalid Ollama API response: ${parseResult.error.message}` };
+    }
+    const models = parseResult.data.models ?? [];
+    const found = models.some(
+      (m) => m.name === cfg.model || m.name.startsWith(`${cfg.model}:`),
+    );
+    if (!found) {
+      const available = models.map((m) => m.name).join(", ") || "none";
+      return {
+        ok: false,
+        error: `Model "${cfg.model}" not found. Available: ${available}. Run: ollama pull ${cfg.model}`,
+      };
+    }
+    return { ok: true };
+  } catch (e: unknown) {
+    const msg = e instanceof Error ? e.message : String(e);
+    return {
+      ok: false,
+      error: `Cannot reach Ollama at ${cfg.baseUrl}: ${msg}. Is Ollama running?`,
+    };
+  }
+}
+
+/**
+ * Embed multiple texts via Ollama /api/embed.
+ * Batches at BATCH_SIZE (Ollama processes sequentially internally).
+ * Returns one Float32Array per input text.
+ */
+export async function embed(
+  texts: string[],
+  config: Partial<OllamaConfig> = {},
+): Promise<Float32Array[]> {
+  const cfg = { ...DEFAULT_CONFIG, ...config };
+  const results: Float32Array[] = [];
+
+  for (let i = 0; i < texts.length; i += BATCH_SIZE) {
+    const batch = texts.slice(i, i + BATCH_SIZE);
+    logDebug(HOOK_NAME, `Embedding batch ${i / BATCH_SIZE + 1} (${batch.length} texts)`);
+
+    const resp = await fetch(`${cfg.baseUrl}/api/embed`, {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ model: cfg.model, input: batch }),
+      signal: AbortSignal.timeout(30_000),
+    });
+
+    if (!resp.ok) {
+      const body = await resp.text().catch(() => "");
+      throw new Error(`Ollama embed failed (${resp.status}): ${body}`);
+    }
+
+    const json = await resp.json();
+    const parseResult = OllamaEmbedResponseSchema.safeParse(json);
+    if (!parseResult.success) {
+      throw new Error(`Invalid Ollama embed response: ${parseResult.error.message}`);
+    }
+
+    const data = parseResult.data;
+    if (data.embeddings.length !== batch.length) {
+      throw new Error(
+        `Expected ${batch.length} embeddings, got ${data.embeddings.length}`,
+      );
+    }
+
+    for (const vec of data.embeddings) {
+      results.push(new Float32Array(vec));
+    }
+  }
+
+  return results;
+}
+
+/**
+ * Embed a single text. Convenience wrapper around embed().
+ */
+export async function embedOne(
+  text: string,
+  config: Partial<OllamaConfig> = {},
+): Promise<Float32Array> {
+  const results = await embed([text], config);
+  if (results.length === 0) {
+    throw new Error("Embedding failed: received empty result array");
+  }
+  return results[0];
+}
+
+/**
+ * Generate text using Ollama's /api/generate endpoint.
+ * Follows same pattern as embed()/embedOne() for consistency.
+ */
+export async function generateText(
+  prompt: string,
+  options?: {
+    systemPrompt?: string;
+    temperature?: number;
+    maxTokens?: number;
+    timeout?: number;
+    model?: string;
+  },
+): Promise<{ success: boolean; text: string; error?: string; latency_ms: number }> {
+  const startTime = Date.now();
+  const model = options?.model ?? HYDE_OLLAMA_MODEL;
+  const baseUrl = process.env.OLLAMA_BASE_URL ?? OLLAMA_BASE_URL;
+
+  try {
+    const response = await fetch(`${baseUrl}/api/generate`, {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({
+        model,
+        prompt: options?.systemPrompt
+          ? `${options.systemPrompt}\n\n${prompt}`
+          : prompt,
+        stream: false, // Non-streaming for simplicity
+        options: {
+          temperature: options?.temperature ?? 0.7,
+          num_predict: options?.maxTokens ?? 200,
+        },
+      }),
+      signal: AbortSignal.timeout(options?.timeout ?? 10_000),
+    });
+
+    if (!response.ok) {
+      return {
+        success: false,
+        text: "",
+        error: `HTTP ${response.status}: ${response.statusText}`,
+        latency_ms: Date.now() - startTime,
+      };
+    }
+
+    const data = await response.json();
+    return {
+      success: true,
+      text: data.response || "",
+      latency_ms: Date.now() - startTime,
+    };
+  } catch (e) {
+    return {
+      success: false,
+      text: "",
+      error: String(e),
+      latency_ms: Date.now() - startTime,
+    };
+  }
+}