tuna-agent 0.1.152 → 0.1.154

package/dist/daemon/analyze-video-handler.js

@@ -247,8 +247,10 @@ Voiceover during this scene: "${voiceoverText || '(none)'}"`;
         { text: promptText },
         ...frames.map(b64 => ({ inlineData: { mimeType: 'image/jpeg', data: b64 } })),
     ];
-    const { text, usage } = await geminiGenerate(parts, 512);
-    cost?.geminiVision('vision', usage);
+    // Gemini 2.5 Flash (cheapest). geminiGenerate disables thinking for
+    // 2.5-flash so the 512 budget isn't starved → non-empty descriptions.
+    const { text, usage } = await geminiGenerate(parts, 512, 'gemini-2.5-flash');
+    cost?.geminiVision('vision', usage, 'gemini-2.5-flash');
     return text;
 }
 // Phase 1 (the strong part of AI_Video_Clone, ported): ONE gpt-4o call over
@@ -269,7 +271,7 @@ async function visionExtractPhase1(frames, transcript, cost) {
         master_cast_prompt: '',
         characters: [],
     };
-    if (!OPENAI_KEY || frames.length === 0)
+    if (!GEMINI_KEYS.length || frames.length === 0)
         return empty;
     try {
         const promptText = `Act as a Master Film Director. These frames are sampled across an ENTIRE video, in order. Use them + the transcript to analyse the whole piece.
@@ -291,25 +293,18 @@ Rules:
 - NO GROUP ENTRIES (CRITICAL): NEVER output a collective/crowd label as a single entry — forbidden: "VILLAGERS", "LADIES GROUP", "KNITTING GROUP", "CROWD", "GROUP OF ...", any "*_GROUP". If 2+ similar secondary people RECUR across scenes, list them as SEPARATE numbered individuals (e.g. WOMAN_1, WOMAN_2, WOMAN_3), each with its OWN distinct face/hair/body/age. Only a truly anonymous one-off background that never recurs may be omitted entirely.
 - characters.description: ENGLISH only, factual, no camera/action words.
 - DISTINCT FACES (CRITICAL): every character MUST have a HIGHLY UNIQUE facial structure, a distinct hairstyle, a specific body type and a clearly different age. NEVER reuse the same or a similar facial description for two characters — they must look completely different from one another.`;
-        // Phase-1 stays on gpt-4o (best cinematic summary + style — the narrative
-        // spine for the cloned script) with a dense 30-frame seed. Only 1
+        // Phase-1 on Gemini 2.5 Flash (cheapest) with a dense 30-frame seed. 1
         // call/video; final cast recall is double-covered by the reconcile pass.
-        const content = [{ type: 'text', text: promptText }];
-        for (const b64 of frames) {
-            content.push({ type: 'image_url', image_url: { url: `data:image/jpeg;base64,${b64}` } });
-        }
-        const res = await fetch('https://api.openai.com/v1/chat/completions', {
-            method: 'POST',
-            headers: { 'Content-Type': 'application/json', Authorization: `Bearer ${OPENAI_KEY}` },
-            body: JSON.stringify({ model: 'gpt-4o', max_tokens: 1600, messages: [{ role: 'user', content }] }),
-        });
-        if (!res.ok)
-            return empty;
-        const data = await res.json();
-        cost?.chat('phase1', 'gpt-4o', data.usage);
-        const rawTxt = (data.choices?.[0]?.message?.content || '').trim();
+        // geminiGenerate disables thinking for 2.5-flash + generous 3000 output
+        // budget so the JSON answer is never starved (was empty without this).
+        const parts = [
+            { text: promptText },
+            ...frames.map(b64 => ({ inlineData: { mimeType: 'image/jpeg', data: b64 } })),
+        ];
+        const { text: rawTxt, usage } = await geminiGenerate(parts, 3000, 'gemini-2.5-flash');
         if (!rawTxt)
             return empty;
+        cost?.geminiVision('phase1', usage, 'gemini-2.5-flash');
         let parsed = {};
         try {
             const m = rawTxt.match(/\{[\s\S]*\}/);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "tuna-agent",
-  "version": "0.1.152",
+  "version": "0.1.154",
   "description": "Tuna Agent - Run AI coding tasks on your machine",
   "bin": {
     "tuna-agent": "dist/cli/index.js"