Use official ollama npm client for all LLM calls

Replaces hand-rolled fetch to /api/chat with the ollama@0.6.3 package.

The package supplies proper Accept/Content-Type/User-Agent headers and

auto-handles the Ollama protocol, fixing the CDN 403s Node's bare fetch

gets against ollama.com during filing.

Both organize and search LLM ranking now route through controller/ollama.js.

Signed-off-by: William Mantly <wmantly@gmail.com>

controller/ollama.js

@@ -1,7 +1,18 @@
 'use strict';
 
+const { Ollama } = require('ollama');
 const conf = require('>/conf');
 
+// Shared Ollama client. The npm package adds its own Accept / Content-Type / User-Agent
+// headers and handles the Ollama chat protocol, which avoids the CDN-level 403s Node's
+// bare fetch gets against ollama.com.
+const client = new Ollama({
+	host: conf.ollama.url,
+	headers: {
+		'Authorization': `Bearer ${conf.ollama.apiKey}`,
+	},
+});
+
 // Pull the first JSON object out of a model response( tolerates ```json fences / prose).
 function extractJSON(content){
 	let text = String(content).replace(/```json/gi, '').replace(/```/g, '').trim();
@@ -19,35 +30,18 @@ function extractJSON(content){
 	}
 }
 
-// One-shot JSON chat against the Ollama cloud model.
+// One-shot JSON chat against the configured Ollama model.
 async function chatJSON(system, user){
-	let res = await fetch(`${conf.ollama.url}/api/chat`, {
-		method: 'POST',
-		headers: {
-			'Content-Type': 'application/json',
-			'Authorization': `Bearer ${conf.ollama.apiKey}`,
-			'User-Agent': 'tpbproxy/1.0 (organize; +https://github.com/wmantly/tpbproxy)',
-		},
-		body: JSON.stringify({
-			model: conf.ollama.model,
-			stream: false,
-			format: 'json',
-			messages: [
-				{ role: 'system', content: system },
-				{ role: 'user', content: user },
-			],
-		}),
+	let res = await client.chat({
+		model: conf.ollama.model,
+		stream: false,
+		format: 'json',
+		messages: [
+			{ role: 'system', content: system },
+			{ role: 'user', content: user },
+		],
 	});
-	if(!res.ok){
-		let body = await res.text().catch(() => '');
-		console.error(`Ollama chat failed (${res.status}):`, body.slice(0, 500));
-		let error = new Error('OllamaError');
-		error.message = `Ollama chat failed( ${res.status})`;
-		error.status = 502;
-		throw error;
-	}
-	let data = await res.json();
-	return extractJSON(data.message && data.message.content);
+	return extractJSON(res.message && res.message.content);
 }
 
-module.exports = { chatJSON, extractJSON };
+module.exports = { client, chatJSON, extractJSON };

controller/search.js

@@ -22,17 +22,6 @@ function humanSize(bytes){
 	return `${bytes.toFixed(bytes < 10 && i > 0 ? 1 : 0)} ${units[i]}`;
 }
 
-async function fetchJSON(url, options){
-	let res = await fetch(url, options);
-	if(!res.ok){
-		let error = new Error(`UpstreamError`);
-		error.message = `Request to ${url.split('?')[0]} failed( ${res.status})`;
-		error.status = 502;
-		throw error;
-	}
-	return await res.json();
-}
-
 // Turn a raw TPB category id into 'movie' | 'tv' | null.
 function videoKind(category){
 	category = parseInt(category, 10);
@@ -179,27 +168,10 @@ async function rankReleases(candidates, meta){
 		imdb: t.imdb || null,
 	}));
 
-	let body = {
-		model: conf.ollama.model,
-		stream: false,
-		format: 'json',
-		messages: [
-			{ role: 'system', content: buildSystemPrompt(meta.title, meta.year, meta.today) },
-			{ role: 'user', content: JSON.stringify(compact) },
-		],
-	};
-
-	let data = await fetchJSON(`${conf.ollama.url}/api/chat`, {
-		method: 'POST',
-		headers: {
-			'Content-Type': 'application/json',
-			'Authorization': `Bearer ${conf.ollama.apiKey}`,
-			'User-Agent': 'tpbproxy/1.0 (search; +https://github.com/wmantly/tpbproxy)',
-		},
-		body: JSON.stringify(body),
-	});
-
-	let parsed = extractJSON(data.message && data.message.content);
+	let parsed = await ollama.chatJSON(
+		buildSystemPrompt(meta.title, meta.year, meta.today),
+		JSON.stringify(compact)
+	);
 	let options = Array.isArray(parsed.options) ? parsed.options : [];
 	if(!options.length) throw new Error('Model returned no options');
 	return options;

package-lock.json

@@ -20,6 +20,7 @@
         "lru-native2": "^1.2.6",
         "moment": "^2.30.1",
         "mustache": "^4.2.0",
+        "ollama": "^0.6.3",
         "p2psub": "^0.1.9",
         "sequelize": "^6.35.2",
         "sequelize-cli": "^6.6.2",
@@ -2933,6 +2934,15 @@
         "url": "https://github.com/sponsors/ljharb"
       }
     },
+    "node_modules/ollama": {
+      "version": "0.6.3",
+      "resolved": "https://registry.npmjs.org/ollama/-/ollama-0.6.3.tgz",
+      "integrity": "sha512-KEWEhIqE5wtfzEIZbDCLH51VFZ6Z3ZSa6sIOg/E/tBV8S51flyqBOXi+bRxlOYKDf8i327zG9eSTb8IJxvm3Zg==",
+      "license": "MIT",
+      "dependencies": {
+        "whatwg-fetch": "^3.6.20"
+      }
+    },
     "node_modules/on-finished": {
       "version": "2.4.1",
       "resolved": "https://registry.npmjs.org/on-finished/-/on-finished-2.4.1.tgz",
@@ -4411,6 +4421,12 @@
         "node": ">=12"
       }
     },
+    "node_modules/whatwg-fetch": {
+      "version": "3.6.20",
+      "resolved": "https://registry.npmjs.org/whatwg-fetch/-/whatwg-fetch-3.6.20.tgz",
+      "integrity": "sha512-EqhiFU6daOA8kpjOWTL0olhVOF3i7OrFzSYiGsEMB8GcXS+RrzauAERX65xMeNWVqxA6HXH2m69Z9LaKKdisfg==",
+      "license": "MIT"
+    },
     "node_modules/whatwg-url": {
       "version": "14.2.0",
       "resolved": "https://registry.npmjs.org/whatwg-url/-/whatwg-url-14.2.0.tgz",

package.json

@@ -23,6 +23,7 @@
     "lru-native2": "^1.2.6",
     "moment": "^2.30.1",
     "mustache": "^4.2.0",
+    "ollama": "^0.6.3",
     "p2psub": "^0.1.9",
     "sequelize": "^6.35.2",
     "sequelize-cli": "^6.6.2",