auto model token limits

tcsenpai · Oct 13, 2024 · 08394ba · 08394ba
1 parent 6a01bb6
commit 08394ba
Show file tree

Hide file tree

Showing 4 changed files with 78 additions and 39 deletions.
diff --git a/background.js b/background.js
@@ -28,7 +28,7 @@ async function summarizeContent(content, systemPrompt) {
   const endpoint = `${
     settings.ollamaEndpoint || "http://localhost:11434"
   }/api/generate`;
-  const model = settings.ollamaModel || "llama2";
+  const model = settings.ollamaModel || "llama3.1:8b";
   const tokenLimit = settings.tokenLimit || 4096;
 
   const maxContentTokens = tokenLimit - estimateTokenCount(systemPrompt) - 100; // Reserve 100 tokens for safety

diff --git a/manifest.json b/manifest.json
@@ -1,38 +1,31 @@
 {
-    "manifest_version": 2,
-    "name": "SpaceLLama",
-    "version": "1.0",
-    "description": "Summarize web pages using OLLAMA",
-    "permissions": [
-      "activeTab",
-      "storage",
-      "<all_urls>",
-      "tabs"
-    ],
-    "browser_action": {
-      "default_title": "SpaceLLama",
-      "default_icon": "icon.png"
-    },
-    "sidebar_action": {
-      "default_title": "SpaceLLama",
-      "default_panel": "sidebar/sidebar.html",
-      "default_icon": "icon.png"
-    },
-    "background": {
-      "scripts": ["background.js"],
-      "persistent": false
-    },
-    "content_scripts": [
-      {
-        "matches": ["<all_urls>"],
-        "js": ["content_scripts/content.js"]
-      }
-    ],
-    "options_ui": {
-      "page": "options/options.html",
-      "open_in_tab": true
-    },
-    "web_accessible_resources": [
-      "sidebar/marked.min.js"
-    ]
-  }
+  "manifest_version": 2,
+  "name": "SpaceLLama",
+  "version": "1.1",
+  "description": "Summarize web pages using Ollama. Supports custom models, token limits, system prompts, chunking, and more. See https://github.com/tcsenpai/spacellama for more information.",
+  "permissions": ["activeTab", "storage", "<all_urls>", "tabs"],
+  "browser_action": {
+    "default_title": "SpaceLLama",
+    "default_icon": "icon.png"
+  },
+  "sidebar_action": {
+    "default_title": "SpaceLLama",
+    "default_panel": "sidebar/sidebar.html",
+    "default_icon": "icon.png"
+  },
+  "background": {
+    "scripts": ["background.js"],
+    "persistent": false
+  },
+  "content_scripts": [
+    {
+      "matches": ["<all_urls>"],
+      "js": ["content_scripts/content.js"]
+    }
+  ],
+  "options_ui": {
+    "page": "options/options.html",
+    "open_in_tab": true
+  },
+  "web_accessible_resources": ["sidebar/marked.min.js", "model_tokens.json"]
+}
diff --git a/model_tokens.json b/model_tokens.json
@@ -0,0 +1,26 @@
+{
+    "llama2": 4096,
+    "llama2:13b": 4096,
+    "llama2:70b": 4096,
+    "codellama": 16384,
+    "codellama:13b": 16384,
+    "codellama:34b": 16384,
+    "mistral": 8192,
+    "mixtral": 32768,
+    "phi": 2048,
+    "qwen": 8192,
+    "qwen:14b": 8192,
+    "qwen:72b": 8192,
+    "stablelm": 4096,
+    "stablelm-zephyr": 4096,
+    "neural-chat": 8192,
+    "openhermes": 8192,
+    "starling-lm": 8192,
+    "orca2": 4096,
+    "vicuna": 8192,
+    "wizardcoder": 16384,
+    "wizardcoder:python": 16384,
+    "wizardmath": 8192,
+    "llama3.1:8b": 128000,
+    "llama3.1:70b": 128000
+  }
diff --git a/options/options.js b/options/options.js
@@ -58,7 +58,9 @@ async function restoreOptions() {
   document.getElementById("endpoint").value = endpoint;
   document.getElementById("model").value = result.ollamaModel || "llama2";
   document.getElementById("system-prompt").value = result.systemPrompt || defaultSystemPrompt;
-  document.getElementById("token-limit").value = result.tokenLimit || 4096;
+
+  await updateTokenLimit();
+
   const isValid = await validateEndpoint(endpoint);
   updateEndpointStatus(isValid);
 }
@@ -72,3 +74,21 @@ document.getElementById("endpoint").addEventListener("blur", async (e) => {
   updateEndpointStatus(isValid);
 });
 
+async function loadModelTokens() {
+  const response = await fetch(browser.runtime.getURL('model_tokens.json'));
+  return await response.json();
+}
+
+async function updateTokenLimit() {
+  const modelTokens = await loadModelTokens();
+  const model = document.getElementById("model").value;
+  const tokenLimitInput = document.getElementById("token-limit");
+
+  if (model in modelTokens) {
+    tokenLimitInput.value = modelTokens[model];
+  } else {
+    tokenLimitInput.value = 4096; // Default value
+  }
+}
+
+document.getElementById("model").addEventListener("change", updateTokenLimit);