opencode-pollinations-plugin 6.1.0-beta.9 → 6.2.1

package/README.md

@@ -1,119 +1,133 @@
-# 🌸 Pollinations AI Plugin for OpenCode
+# 🌸 Pollinations AI Plugin for OpenCode (v6.2.0)
 
 <div align="center">
-  <img src="https://avatars.githubusercontent.com/u/88394740?s=400&v=4" alt="Pollinations.ai Logo" width="150">
+  <img src="https://avatars.githubusercontent.com/u/88394740?s=400&v=4" alt="Pollinations.ai Logo" width="200">
   <br>
-  <b>The Bridge between OpenCode and the Pollinations.ai Ecosystem</b>
+  <b>The Bridge between OpenCode and the Pollinations.ai Ecosystem.</b>
   <br>
-  Access 25+ AI models with Vision, Reasoning, and Tools support.
+  Access a continuous universe of free basic AI models, or leverage premium enterprise models with our generous Daily Free Tiers directly from your editor.
 </div>
 
 <div align="center">
 
-![Version](https://img.shields.io/badge/version-6.0.0--beta.25-blue.svg)
+![Version](https://img.shields.io/badge/version-v6.2.0-blue.svg)
 ![License](https://img.shields.io/badge/license-MIT-green.svg)
-![Status](https://img.shields.io/badge/status-Stable-brightgreen.svg)
-
-[📜 Changelog](./CHANGELOG.md) | [🛣️ Roadmap](./ROADMAP.md)
+![Status](https://img.shields.io/badge/status-Stable-success.svg)
 
 </div>
 
----
+## 📖 Philosophy: Open AI for Creators
 
-## ✨ Features
+> **"No closed doors, no corporate hoops — just good tools and good people."**
 
-- **👁️ Vision Support**: Send images to AI models (OpenAI, Claude, Gemini, Kimi)
-- **🧠 Reasoning Models**: Access advanced thinking models (DeepSeek, Kimi, Perplexity)
-- **🛠️ Native Tools**: Full function calling support for compatible models
-- **🔐 Secure Auth**: Native OpenCode `/connect` integration
-- **🔄 Hot-Reload**: Change API keys without restarting OpenCode
+Pollinations.ai is an open-source platform created by and for the community. We offer a unified, direct API for generating images, text, audio, and video.
+- **Transparent**: Our code, roadmap, and discussions are public.
+- **Fair Economy**: A single currency (**Pollen**) for all media and models. Predictable and transparent pricing.
 
 ---
 
-## 🐝 Understanding Pollen & Tiers
+## ✨ What's new in V6.2? (The Anti-Hallucination Update)
+- **100% Dynamic Engine**: Hardcoded model lists, default configurations, and fixed prices are gone! In V6.2, OpenCode's AI agent now fetches the latest LLM models, parameters, tags (`[💎 Paid]`, `[🌿 Free]`, limits), and cost approximations dynamically from the Pollinations APIs.
+- **Robust Security**: Protection against path traversal and strict URL verifications are fully integrated.
+- **Improved Web Search**: The `polli_web_search` component maps seamlessly to current web-enabled and specialized groundings options like Google Gemini Fast, Perplexity, and Custom specialized assistants.
 
-**Pollen** is Pollinations' unified credit system. $1 ≈ 1 Pollen.
-
-| Tier | Daily Grant | Requirement |
-|:-----|:------------|:------------|
-| 🦠 **Microbe** | 0.1 Pollen | Flagged accounts |
-| 🌱 **Spore** | 1 Pollen | Sign up |
-| 🌿 **Seed** | 3 Pollen | Active GitHub dev (8+ points) |
-| 🌸 **Flower** | 10 Pollen | Published app |
-| 🍯 **Nectar** | 20 Pollen | Major contributor |
+---
 
-> 💡 This plugin was published to the OpenCode ecosystem, granting its author **Flower** tier (10 Pollen/day).
+## 🧰 Tools & Commands V6.2
+
+Beyond text discussion, connecting your key gives OpenCode Agents access to our AI Media Tools powered by Pollinations models:
+
+### 💎 Integrated Generative Tools (ENTER ONLY - requires API key)
+- `polli_gen_image` : State-of-the-art imagery models (`Flux`, `Sana`, `Gemini`).
+- `polli_gen_video` : Powerful Text-to-Video and Image-to-Video capabilities (`Wan`, `Veo`, `LTX`, `Reveal`).
+- `polli_gen_audio` & `polli_gen_music` : Magical voice synthesis (ElevenLabs, TTS) and Generative Music.
+- `polli_stt` : High-flying voice transcription (Whisper V3).
+- `polli_web_search` : Connected Web & Specialized Search context (`gemini-search`, `perplexity...`).
+
+### 🧰 Free Creator Bonus Tools (Always available)
+- `remove_background` : Built-in ultra-fast image background removal.
+- `gen_qrcode`, `extract_frames`, `extract_audio` : Utilities.
+
+### 💻 Complete List of Terminal Commands
+Use the alias **`/poll`** or **`/pollinations`**.
+- `/poll help` : Displays the interactive help table.
+- `/poll connect` : Bring Your Own Key configuration tool (Interactive).
+- `/poll usage full` : Real-time dashboard (Stats), active Freetiers, and Wallet Balance.
+- `/poll config` : Finely adjust Cost Guards, Logs, Language, and Display.
+- `/poll models` : Check the status of available Models.
+- `/poll pricing` : View real-time unified pricing (Average Cost Estimate).
+- `/poll fallback` : Define the ultimate Safety Net Chat model.
+- `/poll mode` : Change mode without going through the API.
+- `/poll infos` : Discover community rules and the leveling system.
+
+### 🛡️ The "Cost Guard" & the "Safety Net"
+We have introduced fundamental protections to ensure your workflow never interrupts and your wallet (Wallet or free tiers) is under your control.
+- **Safety Net**: If you use premium models and your daily Pollen quota runs out in the middle of a chat session, the plugin silently and automatically switches to a free model. *No more blocking errors (429).*
+- **Cost Guard for Tools**: OpenCode Agents can be zealous. If an Agent tries to spend too many Pollens to generate a heavy video or music, the plugin intercepts the request. We have implemented an asynchronous flow that asks for your manual confirmation before executing costly generations. You stay in control.
+
+### 🌍 Native Multilingual Support (i18n)
+Pollinations for OpenCode natively speaks your language.
+- The Engine Interface, Notifications (Toasts), Tool Returns, and Commands are fully translated into **English (Default)**, **French**, **Spanish**, **German**, and **Italian**.
+- Type `/poll config lang <fr|es|de|it>` in the terminal to switch instantly.
 
 ---
 
-## 🚀 Quick Start
+## 🐝 Understanding Pollens & "Free Tiers"
 
-### 1. Install Plugin
-```bash
-npm install -g opencode-pollinations-plugin
-npx opencode-pollinations-plugin
-```
+In the past, Pollinations mainly relied on ad-funded network traffic. Today, running massive models (like Claude 4.5, Flux Pro, Wan Video) costs money. Pollinations therefore introduces the **Enter Universe** which requires an API key and unlocks cutting-edge models.
 
-### 2. Connect Your Account
-```
-/connect
-```
-Select **pollinations** → Enter your API key from [enter.pollinations.ai](https://enter.pollinations.ai)
+**But wait, you don't need a credit card!**
 
-### 3. Start Using
-Select any `pollinations/*` model in OpenCode and chat!
+**Pollen** is our unified credit system ($1 ≈ 1 Pollen). By connecting a simple Free API Key, you unlock daily Pollen reloads according to your Developer Tier:
 
----
+| Tier | Daily Reload | Condition |
+| :--- | :--- | :--- |
+| 🦠 **Microbe** | **0.1 Pollen/day** | Just register! |
+| 🍄 **Spore** | **1 Pollen/day** | Automatic verification |
+| 🌱 **Seed** | **3 Pollen/day** | Active GitHub Developer (8+ points) |
+| 🌸 **Flower** | **10 Pollen/day** | **Publish an App** (Like this Plugin!) |
 
-## 👁️ Vision Support
+> 🎁 **Get your Free Personal Key (BYOK) on [Pollinations.ai](https://enter.pollinations.ai/authorize?redirect_url=https://github.com/fkom13/opencode-pollinations-plugin) to boost OpenCode!**
 
-The following models support image input:
+*(Note: We still maintain the "Free Universe" fallback for basic chat (`openai-fast`) which requires no key, but its capacity is very limited and mainly designed as a safety net).*
 
-| Model | Vision | Reasoning | Tools |
-|-------|:------:|:---------:|:-----:|
-| `openai` / `openai-large` | ✅ | - | ✅ |
-| `gemini` / `gemini-search` | ✅ | - | ✅ |
-| `claude` / `claude-large` | ✅ | - | ✅ |
-| `kimi` | ✅ | ✅ | - |
-| `openai-audio` | ✅ | - | ✅ |
+Paid pollens allow you to access even more powerful and premium models.
 
-**Usage**: Simply paste an image in the chat or use an image URL. The plugin handles encoding automatically.
+Daily free tier pollen credits are consumed before touching the wallet (purchased pollen) except for paid models.
 
 ---
 
-## 📊 Available Models (25+)
+## 🚀 Getting Started & Onboarding
 
-### Text/Chat Models
-- **OpenAI**: `openai`, `openai-fast`, `openai-large` 
-- **Gemini**: `gemini`, `gemini-fast`, `gemini-search`, `gemini-large` 💎
-- **Claude**: `claude-fast`, `claude`, `claude-large` 💎
-- **Reasoning**: `deepseek`, `kimi`, `perplexity-reasoning`
-- **Code**: `qwen-coder`, `mistral`
-- **Fast**: `nova-fast`, `grok`
+### 🐧 1. Cross-Platform Configuration (NPM Installation)
+This plugin is **fully cross-platform** (Windows, macOS, Linux) and detects its ports dynamically.
 
-### Paid-Only Models (💎)
-These require purchased credits (not daily grants):
-- `claude-large`, `gemini-large`, `veo`, `seedream-pro`
+1. Global installation:
+   ```bash
+   npm install -g opencode-pollinations-plugin
+   ```
+2. Auto-Configuration:
+   ```bash
+   npx opencode-pollinations-plugin
+   ```
+   *(Or inject it manually into `~/.config/opencode/opencode.json`)*
 
----
+### 🔑 2. Interactive Onboarding
+Once in OpenCode, simply type the following command in the Agent Terminal:
+```bash
+/poll connect
+```
+An interactive conversational assistant will guide you to inject your Pollinations Key and configure your space. *Restart OpenCode to update the list of models in the UI interface.*
 
-## 🔧 Commands
 
-```
-/pollinations usage       # View Pollen balance
-/pollinations usage full  # Detailed model breakdown
-/pollinations status      # Plugin health check
-```
 
 ---
 
 ## 🔗 Links
 
-- **Get API Key**: [enter.pollinations.ai](https://enter.pollinations.ai)
-- **Discord**: [Join Community](https://discord.gg/pollinations-ai-885844321461485618)
-- **GitHub**: [Plugin Repository](https://github.com/fkom13/opencode-pollinations-plugin)
-
----
+- **Create your Pollen API Key**: [pollinations.ai](https://pollinations.ai)
+- **Discord Community**: [Join us!](https://discord.gg/pollinations-ai-885844321461485618)
+- **OpenCode Ecosystem**: [opencode.ai](https://opencode.ai/docs/ecosystem#plugins)
 
 ## 📜 License
 

package/dist/index.js

@@ -1,31 +1,21 @@
 import * as http from 'http';
-import * as fs from 'fs';
 import { generatePollinationsConfig } from './server/generate-config.js';
-import { loadConfig, saveConfig } from './server/config.js';
+import { loadConfig, migrateLegacyConfig } from './server/config.js';
 import { handleChatCompletion } from './server/proxy.js';
-import { createToastHooks, setGlobalClient } from './server/toast.js';
+import { createToastHooks, createToolHooks, setGlobalClient } from './server/toast.js';
 import { createStatusHooks } from './server/status.js';
-import { createCommandHooks } from './server/commands.js';
+import { createCommandHooks, setClientForCommands } from './server/commands.js';
+import { createToolRegistry } from './tools/index.js';
 import { createRequire } from 'module';
 const require = createRequire(import.meta.url);
-const LOG_FILE = '/tmp/opencode_pollinations_v4.log';
-function log(msg) {
-    try {
-        fs.appendFileSync(LOG_FILE, `[${new Date().toISOString()}] ${msg}\n`);
-    }
-    catch (e) { }
-}
-// === PROXY SERVER (Singleton with Fixed Port) ===
-const DEFAULT_PORT = 18888;
-const GLOBAL_SERVER_KEY = '__POLLINATIONS_PROXY_SERVER__';
+import { log } from './server/logger.js';
+import { ModelRegistry } from './server/models/index.js';
+const sessionModels = new Map();
 const startProxy = () => {
-    // Check if server exists in global scope (survives module reloads)
-    if (global[GLOBAL_SERVER_KEY]) {
-        log(`[Proxy] Reusing existing global server on port ${DEFAULT_PORT}`);
-        return Promise.resolve(DEFAULT_PORT);
-    }
     return new Promise((resolve) => {
         const server = http.createServer(async (req, res) => {
+            // ... (Request Handling) ...
+            // We reuse the existing logic structure but simplified startup
             log(`[Proxy] Request: ${req.method} ${req.url}`);
             res.setHeader('Access-Control-Allow-Origin', '*');
             res.setHeader('Access-Control-Allow-Methods', 'GET, POST, OPTIONS');
@@ -67,156 +57,58 @@ const startProxy = () => {
             res.writeHead(404);
             res.end("Not Found");
         });
-        // Try fixed port first, fallback to dynamic if occupied
-        const tryListen = (port, fallbackToDynamic) => {
-            server.listen(port, '127.0.0.1', () => {
-                // @ts-ignore
-                const assignedPort = server.address().port;
-                global[GLOBAL_SERVER_KEY] = server;
-                log(`[Proxy] Started v${require('../package.json').version} on port ${assignedPort}${port === 0 ? ' (dynamic fallback)' : ''}`);
-                resolve(assignedPort);
-            });
-            server.on('error', (e) => {
-                if (e.code === 'EADDRINUSE' && fallbackToDynamic) {
-                    log(`[Proxy] Port ${port} in use, falling back to dynamic port`);
-                    server.removeAllListeners('error');
-                    tryListen(0, false); // Try dynamic port
-                }
-                else {
-                    log(`[Proxy] Fatal Error: ${e}`);
-                    resolve(0);
-                }
-            });
-        };
-        tryListen(DEFAULT_PORT, true);
+        // Listen on random port (0) to avoid conflicts (CLI/IDE)
+        server.listen(0, '127.0.0.1', () => {
+            // @ts-ignore
+            const assignedPort = server.address().port;
+            log(`[Proxy] Started v${require('../package.json').version} (Dynamic Port) on port ${assignedPort}`);
+            resolve(assignedPort);
+        });
+        server.on('error', (e) => {
+            log(`[Proxy] Fatal Error: ${e}`);
+            resolve(0);
+        });
     });
 };
-// === AUTH HOOK: Native /connect Integration ===
-// Auth Hook moved inside plugin to access context
 // === PLUGIN EXPORT ===
 export const PollinationsPlugin = async (ctx) => {
-    log(`Plugin Initializing v${require('../package.json').version}...`);
-    // START PROXY on fixed port
+    const v = require('../package.json').version;
+    log(`Plugin Initializing v${v}...`);
+    log(`[ENV] Keys: ${Object.keys(process.env).filter(k => k.includes('OPENCODE') || k.includes('APP') || k.includes('DATA')).join(', ')}`);
+    // MIGRATE CONFIG
+    migrateLegacyConfig();
+    // START PROXY
     const port = await startProxy();
     const localBaseUrl = `http://127.0.0.1:${port}/v1`;
+    // INIT MODEL REGISTRY (non-blocking, fire-and-forget)
+    ModelRegistry.refresh().then(() => {
+        const stats = ModelRegistry.stats();
+        log(`[ModelRegistry] Ready: ${stats.image} image, ${stats.video} video, ${stats.audio} audio, ${stats.text} text`);
+        // Démarrage du patcher asynchrone des descriptions des Outils (Phase 1.5)
+        import('./server/models/worker.js').then(module => {
+            module.ToolRegistryWorker.start();
+        }).catch(e => log(`[ToolWorker] Failed to load worker: ${e}`));
+    }).catch(e => log(`[ModelRegistry] Init failed (will use fallback): ${e}`));
     setGlobalClient(ctx.client);
+    setClientForCommands(ctx.client);
     const toastHooks = createToastHooks(ctx.client);
     const commandHooks = createCommandHooks();
-    // Helper: Refresh provider config (for hot-reload after /connect)
-    let isRefreshing = false;
-    const refreshProviderConfig = async () => {
-        if (isRefreshing)
-            return;
-        isRefreshing = true;
-        try {
-            log('[Event] Refreshing provider config after auth update...');
-            const modelsArray = await generatePollinationsConfig();
-            const modelsObj = {};
-            for (const m of modelsArray) {
-                modelsObj[m.id] = m;
-            }
-            const version = require('../package.json').version;
-            // CRITICAL: Fetch current config first to avoid overwriting other providers
-            let currentConfig = {};
-            try {
-                // Try to fetch existing config to preserve other providers
-                const response = await ctx.client.fetch('/config');
-                if (response.ok) {
-                    currentConfig = await response.json();
-                }
-            }
-            catch (err) {
-                log(`[Event] Warning: Could not fetch current config: ${err}`);
-            }
-            // Safe Merge
-            if (!currentConfig.provider)
-                currentConfig.provider = {};
-            currentConfig.provider.pollinations = {
-                id: 'openai',
-                name: `Pollinations AI (v${version})`,
-                options: {
-                    baseURL: localBaseUrl,
-                    apiKey: 'plugin-managed',
-                },
-                models: modelsObj
-            };
-            // Use Server API to update config with the MERGED object
-            await ctx.client.fetch('/config', {
-                method: 'PATCH',
-                headers: { 'Content-Type': 'application/json' },
-                body: JSON.stringify({
-                    provider: currentConfig.provider
-                })
-            });
-            log(`[Event] Provider config refreshed with ${Object.keys(modelsObj).length} models.`);
-        }
-        catch (e) {
-            log(`[Event] Failed to refresh provider config: ${e}`);
-        }
-        finally {
-            // Debounce: prevent another refresh for 5 seconds
-            setTimeout(() => { isRefreshing = false; }, 5000);
-        }
-    };
+    // Build tool registry (conditional on API key presence)
+    const toolRegistry = createToolRegistry();
+    log(`[Tools] ${Object.keys(toolRegistry).length} tools registered`);
     return {
-        // AUTH HOOK: Native /connect integration (INLINED)
-        auth: {
-            provider: 'pollinations',
-            loader: async (auth, provider) => {
-                log('[AuthHook] loader() called');
-                try {
-                    const authData = await auth();
-                    if (authData && 'key' in authData) {
-                        const k = authData.key;
-                        saveConfig({ apiKey: k });
-                        return { apiKey: k };
-                    }
-                }
-                catch (e) {
-                    log(`[AuthHook] loader error: ${e}`);
-                }
-                const config = loadConfig();
-                if (config.apiKey)
-                    return { apiKey: config.apiKey };
-                return {};
-            },
-            methods: [{
-                    type: 'api',
-                    label: 'Pollinations API Key',
-                    prompts: [{
-                            type: 'text',
-                            key: 'apiKey',
-                            message: 'Enter Pollinations API Key (starts with sk_)',
-                            validate: (v) => (!v || v.length < 10) ? 'Invalid key' : undefined
-                        }],
-                    authorize: async (inputs) => {
-                        log(`[AuthHook] authorize() called`);
-                        if (!inputs?.apiKey)
-                            return { type: 'failed' };
-                        try {
-                            const r = await fetch('https://gen.pollinations.ai/text/models', {
-                                headers: { 'Authorization': `Bearer ${inputs.apiKey}` }
-                            });
-                            if (r.ok) {
-                                log('[AuthHook] Success. Saving & Refreshing Config...');
-                                saveConfig({ apiKey: inputs.apiKey });
-                                // CRITICAL: Refresh config IMMEDIATELY after successful auth
-                                await refreshProviderConfig();
-                                return { type: 'success', key: inputs.apiKey };
-                            }
-                        }
-                        catch (e) {
-                            log(`[AuthHook] Auth error: ${e}`);
-                        }
-                        return { type: 'failed' };
-                    }
-                }]
+        "chat.message": async (input) => {
+            const m = input.model;
+            if (m && m.modelID) {
+                sessionModels.set(input.sessionID, `${m.providerID}/${m.modelID}`);
+                log(`[Hook] Saved active model ${m.providerID}/${m.modelID} for session ${input.sessionID}`);
+            }
         },
-        // Event hook removed (logic moved to authorize)
-        event: async ({ event }) => { },
+        tool: toolRegistry,
         async config(config) {
             log("[Hook] config() called");
-            // Generate models based on current auth state
+            // STARTUP only - No complex hot reload logic
+            // The user must restart OpenCode to refresh this list if they change keys.
             const modelsArray = await generatePollinationsConfig();
             const modelsObj = {};
             for (const m of modelsArray) {
@@ -224,19 +116,18 @@ export const PollinationsPlugin = async (ctx) => {
             }
             if (!config.provider)
                 config.provider = {};
+            // Dynamic Provider Name
             const version = require('../package.json').version;
             config.provider['pollinations'] = {
-                id: 'openai',
+                id: 'pollinations',
                 name: `Pollinations AI (v${version})`,
-                options: {
-                    baseURL: localBaseUrl,
-                    apiKey: 'plugin-managed', // Key is managed by auth hook
-                },
+                options: { baseURL: localBaseUrl },
                 models: modelsObj
             };
             log(`[Hook] Registered ${Object.keys(modelsObj).length} models.`);
         },
         ...toastHooks,
+        ...createToolHooks(ctx.client),
         ...createStatusHooks(ctx.client),
         ...commandHooks
     };