fix: custom fetch wrapper for ollama inference. Fixes #656 (#1032)

* Add configurable fetch timeout for Ollama client

* Worker service needs access to the .env file

* repair typescript types

* Update customFetch.ts

* update the config docs

---------

Co-authored-by: sbarbett <shane@barbetta.me>
Co-authored-by: Mohamed Bassem <me@mbassem.com>

3 files changed, 17 insertions, 0 deletions

diff --git a/packages/shared/config.ts b/packages/shared/config.ts
index adc5749e..6e5a4404 100644
--- a/packages/shared/config.ts
+++ b/packages/shared/config.ts
@@ -22,6 +22,7 @@ const allEnv = z.object({
   OLLAMA_BASE_URL: z.string().url().optional(),
   OLLAMA_KEEP_ALIVE: z.string().optional(),
   INFERENCE_JOB_TIMEOUT_SEC: z.coerce.number().default(30),
+  INFERENCE_FETCH_TIMEOUT_SEC: z.coerce.number().default(300),
   INFERENCE_TEXT_MODEL: z.string().default("gpt-4o-mini"),
   INFERENCE_IMAGE_MODEL: z.string().default("gpt-4o-mini"),
   EMBEDDING_TEXT_MODEL: z.string().default("text-embedding-3-small"),
@@ -84,6 +85,7 @@ const serverConfigSchema = allEnv.transform((val) => {
     },
     inference: {
       jobTimeoutSec: val.INFERENCE_JOB_TIMEOUT_SEC,
+      fetchTimeoutSec: val.INFERENCE_FETCH_TIMEOUT_SEC,
       openAIApiKey: val.OPENAI_API_KEY,
       openAIBaseUrl: val.OPENAI_BASE_URL,
       ollamaBaseUrl: val.OLLAMA_BASE_URL,
diff --git a/packages/shared/customFetch.ts b/packages/shared/customFetch.ts
new file mode 100644
index 00000000..e9ac8a89
--- /dev/null
+++ b/packages/shared/customFetch.ts
@@ -0,0 +1,13 @@
+import serverConfig from "./config";
+
+// Custom fetch function with configurable timeout
+export function customFetch(
+  input: Parameters<typeof fetch>[0],
+  init?: Parameters<typeof fetch>[1],
+): ReturnType<typeof fetch> {
+  const timeout = serverConfig.inference.fetchTimeoutSec * 1000; // Convert to milliseconds
+  return fetch(input, {
+    signal: AbortSignal.timeout(timeout),
+    ...init,
+  });
+}
diff --git a/packages/shared/inference.ts b/packages/shared/inference.ts
index e5ddf5ca..92d9dd94 100644
--- a/packages/shared/inference.ts
+++ b/packages/shared/inference.ts
@@ -2,6 +2,7 @@ import { Ollama } from "ollama";
 import OpenAI from "openai";
 
 import serverConfig from "./config";
+import { customFetch } from "./customFetch";
 import logger from "./logger";
 
 export interface InferenceResponse {
@@ -153,6 +154,7 @@ class OllamaInferenceClient implements InferenceClient {
   constructor() {
     this.ollama = new Ollama({
       host: serverConfig.inference.ollamaBaseUrl,
+      fetch: customFetch, // Use the custom fetch with configurable timeout
     });
   }