lmlite 0.2.1__py3-none-any.whl

lmlite/__init__.py

@@ -0,0 +1,8 @@
+# Copyright 2026 Emmanuel Decitre
+# SPDX-License-Identifier: Apache-2.0
+
+from .llm import LLM
+
+__version__ = "0.2.1"
+
+__all__ = ["LLM", "__version__"]

lmlite/bridge/runtime.py

@@ -0,0 +1,27 @@
+# Copyright 2026 Emmanuel Decitre
+# SPDX-License-Identifier: Apache-2.0
+
+try:
+    from js import eval as js_eval
+except ImportError:
+
+    def js_eval(x):
+        return None
+
+
+_runtime_loaded = False
+
+
+def ensure_runtime():
+    global _runtime_loaded
+
+    if _runtime_loaded:
+        return
+
+    from importlib.resources import files
+
+    js_code = files("lmlite").joinpath("js/llm.js").read_text()
+
+    js_eval(js_code)
+
+    _runtime_loaded = True

lmlite/js/llm.js

@@ -0,0 +1,489 @@
+// lmLite JS runtime
+
+class BrowserLLM {
+  constructor(config) {
+    this.generatorModel = config.generator_model;
+    this.embeddingModel = config.embedding_model;
+    this.useLocalModels = config.use_local_models ?? false;
+    this.localModelsPath = config.local_models_path ?? '/drive/models';
+    this.emfsModelUri = config.emfs_model_uri ?? null;
+
+    this.max_new_tokens = config.max_new_tokens ?? 50;
+    this.temperature = config.temperature ?? 0.7;
+    this.top_k = config.top_k ?? 50;
+    this.do_sample = config.do_sample ?? true;
+
+    this.generator = null;
+    this.embedder = null;
+  }
+
+  async init() {
+    const { pipeline, env } = await import(
+      "https://cdn.jsdelivr.net/npm/@xenova/transformers"
+    );
+
+    // Configure for browser environment
+    if (this.useLocalModels) {
+      env.allowLocalModels = false;
+      env.useBrowserCache = false;
+      env.allowRemoteModels = true; // Keep true but intercept fetch
+
+      // Set up custom backend to read from Pyodide FS
+      const pyodide = globalThis.pyodide;
+      if (pyodide) {
+        env.backends.onnx.wasm.proxy = false; // Disable web worker for local files
+
+        // Custom fetch that reads from Pyodide FS
+        const originalFetch = globalThis.fetch;
+        globalThis.fetch = async (url, options) => {
+          const urlStr = typeof url === 'string' ? url : url.toString();
+
+          // Check if this is a model file request for HuggingFace
+          if (urlStr.includes('huggingface.co') &&
+              (urlStr.includes(`Xenova/${this.generatorModel}`) ||
+               (this.embeddingModel && urlStr.includes(`Xenova/${this.embeddingModel}`)))) {
+
+            // Extract filename from URL
+            const parts = urlStr.split('/');
+            const filename = parts[parts.length - 1];
+            const modelName = parts[parts.indexOf('Xenova') + 1];
+            const filepath = `${this.localModelsPath}/Xenova/${modelName}/${filename}`;
+
+            console.log(`Intercepting HF request, reading from Pyodide FS: ${filepath}`);
+
+            try {
+              const data = pyodide.FS.readFile(filepath);
+              const blob = new Blob([data]);
+
+              return new Response(blob, {
+                status: 200,
+                statusText: 'OK',
+                headers: new Headers({
+                  'Content-Type': 'application/octet-stream',
+                  'Content-Length': data.length.toString()
+                })
+              });
+            } catch (err) {
+              console.error(`Failed to read ${filepath}:`, err);
+              return new Response(null, { status: 404, statusText: 'Not Found' });
+            }
+          }
+
+          // Fall back to original fetch for non-model requests
+          return originalFetch(url, options);
+        };
+
+        console.log(`Using local models from Pyodide FS: ${this.localModelsPath}`);
+      }
+    } else {
+      env.allowLocalModels = false;
+      env.useBrowserCache = true;
+      console.log('Using remote models with browser cache (IndexedDB)');
+    }
+
+    try {
+      let modelPath;
+      if (this.useLocalModels) {
+        // Use HuggingFace path format - our custom fetch will intercept
+        modelPath = `Xenova/${this.generatorModel}`;
+      } else {
+        // Download from HuggingFace
+        modelPath = `Xenova/${this.generatorModel}`;
+      }
+
+      console.log(`Loading text-generation model: ${modelPath}`);
+
+      // Add progress callback
+      this.generator = await pipeline(
+        "text-generation",
+        modelPath,
+        {
+          // Only use quantized models when downloading from HuggingFace
+          // Local models are usually not quantized
+          quantized: !this.useLocalModels,
+          // Explicitly use WASM backend
+          device: 'wasm',
+          // Progress callback
+          progress_callback: (progress) => {
+            if (progress.status === 'downloading') {
+              console.log(`Downloading ${progress.file}: ${progress.progress?.toFixed(2)}%`);
+            } else if (progress.status === 'done') {
+              console.log(`Downloaded ${progress.file}`);
+            } else if (progress.status === 'ready') {
+              console.log(`Model ready: ${progress.file}`);
+            } else if (progress.status === 'progress') {
+              console.log(`Loading ${progress.file}: ${progress.progress?.toFixed(2)}%`);
+            }
+          }
+        }
+      );
+
+      console.log(`Model loaded successfully: ${this.generatorModel}`);
+
+      // Warm up the model with a dummy inference to ensure it's fully ready
+      console.log(`Warming up model...`);
+      await this.generator("test", { max_new_tokens: 1 });
+      console.log(`Model warm-up complete`);
+
+    } catch (error) {
+      throw new Error(
+        `Failed to load text-generation model '${this.generatorModel}': ${error.message}. ` +
+        `This may be due to model incompatibility with browser WASM/WebGPU. ` +
+        `Try a different model like 'gpt2' or check model availability.`
+      );
+    }
+
+    if (this.embeddingModel) {
+      try {
+        // Always use HuggingFace format - custom fetch intercepts if local
+        const embeddingPath = `Xenova/${this.embeddingModel}`;
+
+        console.log(`Loading embedding model: ${embeddingPath}`);
+        this.embedder = await pipeline(
+          "feature-extraction",
+          embeddingPath
+        );
+        console.log(`Embedding model loaded successfully: ${this.embeddingModel}`);
+      } catch (error) {
+        throw new Error(
+          `Failed to load embedding model '${this.embeddingModel}': ${error.message}`
+        );
+      }
+    }
+  }
+
+  async exportModelFiles(modelName) {
+    try {
+      console.log(`Starting export for model: ${modelName}`);
+
+      // Check both Cache API and IndexedDB
+      const cacheNames = await caches.keys();
+      console.log('Available Cache API caches:', cacheNames);
+
+      // List all IndexedDB databases to find the right one
+      const databases = await indexedDB.databases();
+      console.log('Available IndexedDB databases:', databases.map(db => db.name));
+
+      // First try Cache API (Transformers.js v3+ uses this)
+      const modelCacheName = cacheNames.find(name =>
+        name.includes('transformers') ||
+        name.includes('huggingface') ||
+        name.includes('onnx')
+      );
+
+      if (modelCacheName) {
+        console.log(`Found model cache: ${modelCacheName}`);
+        return await this._exportFromCacheAPI(modelName, modelCacheName);
+      }
+
+      // Try different possible database names for IndexedDB (older versions)
+      const possibleDbNames = [
+        'transformers-cache',
+        'transformers_cache',
+        'huggingface-assets',
+        'transformers.js',
+        ...databases.map(db => db.name)
+      ];
+
+      let db = null;
+      let dbName = null;
+
+      for (const name of possibleDbNames) {
+        try {
+          db = await new Promise((resolve, reject) => {
+            const request = indexedDB.open(name);
+            request.onsuccess = () => {
+              if (request.result.objectStoreNames.length > 0) {
+                resolve(request.result);
+              } else {
+                request.result.close();
+                resolve(null);
+              }
+            };
+            request.onerror = () => resolve(null);
+          });
+          if (db) {
+            dbName = name;
+            console.log(`Found database: ${dbName}`);
+            break;
+          }
+        } catch (e) {
+          console.log(`Database ${name} not found or empty`);
+        }
+      }
+
+      if (!db) {
+        console.error('No IndexedDB cache found. Available databases:', databases);
+        return `Error: No model cache database found. Models may still be downloading or using a different cache mechanism.`;
+      }
+
+      console.log(`Object stores in ${dbName}:`, Array.from(db.objectStoreNames));
+
+      const storeName = db.objectStoreNames[0];
+      const transaction = db.transaction([storeName], 'readonly');
+      const store = transaction.objectStore(storeName);
+
+      const allKeys = await new Promise((resolve, reject) => {
+        const request = store.getAllKeys();
+        request.onsuccess = () => resolve(request.result);
+        request.onerror = () => reject(request.error);
+      });
+
+      console.log(`Total cached entries: ${allKeys.length}`);
+      console.log(`Sample keys:`, allKeys.slice(0, 5));
+
+      // Filter keys for this model
+      const modelKeys = allKeys.filter(key =>
+        String(key).includes(`Xenova/${modelName}`) || String(key).includes(modelName)
+      );
+
+      console.log(`Found ${modelKeys.length} cached files for ${modelName}`);
+
+      if (modelKeys.length === 0) {
+        return `Error: No cached files found for model '${modelName}'. Model may not be cached yet.`;
+      }
+
+      // Create directory for model files
+      const pyodide = globalThis.pyodide;
+      if (!pyodide) {
+        return 'Error: Pyodide not available, cannot export to filesystem';
+      }
+
+      const modelDir = `/drive/models/Xenova/${modelName}`;
+      pyodide.FS.mkdirTree(modelDir);
+      console.log(`Created directory: ${modelDir}`);
+
+      let exportedCount = 0;
+
+      // Export each file
+      for (const key of modelKeys) {
+        const data = await new Promise((resolve, reject) => {
+          const request = store.get(key);
+          request.onsuccess = () => resolve(request.result);
+          request.onerror = () => reject(request.error);
+        });
+
+        console.log(`Processing key: ${key}, data type:`, typeof data, data);
+
+        if (data) {
+          let arrayBuffer;
+
+          // Handle different data formats
+          if (data instanceof Blob) {
+            arrayBuffer = await data.arrayBuffer();
+          } else if (data.blob instanceof Blob) {
+            arrayBuffer = await data.blob.arrayBuffer();
+          } else if (data instanceof ArrayBuffer) {
+            arrayBuffer = data;
+          } else if (data.data instanceof ArrayBuffer) {
+            arrayBuffer = data.data;
+          } else {
+            console.warn(`Unknown data format for key ${key}:`, data);
+            continue;
+          }
+
+          const uint8Array = new Uint8Array(arrayBuffer);
+
+          // Extract filename from key
+          const filename = String(key).split('/').pop();
+          const filepath = `${modelDir}/${filename}`;
+
+          pyodide.FS.writeFile(filepath, uint8Array);
+          console.log(`Exported: ${filepath} (${uint8Array.length} bytes)`);
+          exportedCount++;
+        }
+      }
+
+      db.close();
+
+      const message = `Successfully exported ${exportedCount} files to ${modelDir}`;
+      console.log(message);
+      return message;
+
+    } catch (error) {
+      const errorMsg = `Export failed: ${error.message}`;
+      console.error(errorMsg, error);
+      return errorMsg;
+    }
+  }
+
+  async _exportFromCacheAPI(modelName, cacheName) {
+    try {
+      const cache = await caches.open(cacheName);
+      const requests = await cache.keys();
+
+      console.log(`Total cached requests: ${requests.length}`);
+
+      // Filter for this model
+      const modelRequests = requests.filter(req =>
+        req.url.includes(`Xenova/${modelName}`) || req.url.includes(modelName)
+      );
+
+      console.log(`Found ${modelRequests.length} cached files for ${modelName}`);
+
+      if (modelRequests.length === 0) {
+        return `Error: No cached files found for model '${modelName}' in Cache API.`;
+      }
+
+      // Access Pyodide filesystem (in web worker context)
+      // Try different ways to access pyodide
+      let pyodide = null;
+      try {
+        pyodide = self.pyodide;
+      } catch (e) {}
+
+      if (!pyodide) {
+        try {
+          pyodide = globalThis.pyodide;
+        } catch (e) {}
+      }
+
+      if (!pyodide) {
+        console.error('Pyodide not found in self or globalThis');
+        // Return data to Python side instead
+        return await this._exportViaPython(modelName, cacheName, modelRequests);
+      }
+
+      console.log('Pyodide found, exporting via FS:', pyodide);
+
+      const modelDir = `/drive/models/Xenova/${modelName}`;
+
+      // Create directory using Pyodide FS
+      try {
+        pyodide.FS.mkdirTree(modelDir);
+        console.log(`Created directory: ${modelDir}`);
+      } catch (e) {
+        console.error('Failed to create directory:', e);
+        return `Error: Failed to create directory ${modelDir}: ${e.message}`;
+      }
+
+      let exportedCount = 0;
+
+      for (const request of modelRequests) {
+        const response = await cache.match(request);
+        if (response) {
+          const arrayBuffer = await response.arrayBuffer();
+          const uint8Array = new Uint8Array(arrayBuffer);
+
+          // Extract filename from URL
+          const url = new URL(request.url);
+          const pathParts = url.pathname.split('/');
+          const filename = pathParts[pathParts.length - 1];
+
+          const filepath = `${modelDir}/${filename}`;
+          pyodide.FS.writeFile(filepath, uint8Array);
+          console.log(`Exported: ${filepath} (${uint8Array.length} bytes)`);
+          exportedCount++;
+        }
+      }
+
+      const message = `Successfully exported ${exportedCount} files to ${modelDir}`;
+      console.log(message);
+      return message;
+
+    } catch (error) {
+      const errorMsg = `Cache API export failed: ${error.message}`;
+      console.error(errorMsg, error);
+      return errorMsg;
+    }
+  }
+
+  async _exportViaPython(modelName, cacheName, modelRequests) {
+    // Export by returning data to Python side
+    console.log('Exporting via Python side');
+
+    const cache = await caches.open(cacheName);
+    const files = [];
+
+    for (const request of modelRequests) {
+      const response = await cache.match(request);
+      if (response) {
+        const arrayBuffer = await response.arrayBuffer();
+        const uint8Array = new Uint8Array(arrayBuffer);
+
+        // Extract filename from URL
+        const url = new URL(request.url);
+        const pathParts = url.pathname.split('/');
+        const filename = pathParts[pathParts.length - 1];
+
+        files.push({
+          filename: filename,
+          data: uint8Array
+        });
+      }
+    }
+
+    return {
+      success: true,
+      model: modelName,
+      files: files
+    };
+  }
+
+  async generate(prompt) {
+    if (!this.generator) {
+      throw new Error("Generator model not initialized. Call init() first.");
+    }
+
+    try {
+      const output = await this.generator(prompt, {
+        max_new_tokens: this.max_new_tokens,
+        temperature: this.temperature,
+        top_k: this.top_k,
+        do_sample: this.do_sample
+      });
+
+      return output[0].generated_text;
+    } catch (error) {
+      throw new Error(
+        `Text generation failed: ${error.message}. ` +
+        `This could be due to: ` +
+        `(1) Model still downloading - wait and retry, ` +
+        `(2) Memory constraints in browser, ` +
+        `(3) ONNX Runtime error - try a different model like 'gpt2', ` +
+        `(4) Model incompatibility with browser WASM environment.`
+      );
+    }
+  }
+
+  async embed(text) {
+    if (!this.embedder) {
+      throw new Error("No embedding model configured");
+    }
+
+    const result = await this.embedder(text);
+    return Array.from(result.data);
+  }
+}
+
+// --- Factory ---
+
+function createLLM(config) {
+  return new BrowserLLM(config);
+}
+
+// --- Similarity helpers ---
+
+function cosineSimilarity(a, b) {
+  let dot = 0.0, normA = 0.0, normB = 0.0;
+
+  for (let i = 0; i < a.length; i++) {
+    dot += a[i] * b[i];
+    normA += a[i] * a[i];
+    normB += b[i] * b[i];
+  }
+  return dot / (Math.sqrt(normA) * Math.sqrt(normB));
+}
+
+function cosineSimilarityBatch(queryVec, docVecs) {
+  return docVecs.map(doc => cosineSimilarity(queryVec, doc));
+}
+
+// --- expose globally (critical for Python) ---
+
+globalThis.createLLM = createLLM;
+globalThis.cosineSimilarity = cosineSimilarity;
+globalThis.cosineSimilarityBatch = cosineSimilarityBatch;
+
+undefined;
+

lmlite/llm.py

@@ -0,0 +1,223 @@
+# Copyright 2026 Emmanuel Decitre
+# SPDX-License-Identifier: Apache-2.0
+
+from .bridge.runtime import ensure_runtime
+
+
+class LLM:
+    DEFAULTS = {
+        "generator_model": "gpt2",
+        "embedding_model": "all-MiniLM-L6-v2",
+        "max_new_tokens": 50,
+        "temperature": 0.7,
+        "top_k": 50,
+        "do_sample": True,
+        "use_local_models": False,
+        "local_models_path": "/drive/models",
+        "auto_detect_local": True,  # Automatically use local models if available
+    }
+
+    def __init__(self, js_instance):
+        self.js = js_instance
+
+    # --- lifecycle ---
+
+    @classmethod
+    async def create(cls, **overrides):
+        import os
+
+        ensure_runtime()
+
+        from js import createLLM
+
+        config = {**cls.DEFAULTS, **overrides}
+
+        # Auto-detect local models if enabled
+        if config.get("auto_detect_local", True) and not config.get("use_local_models", False):
+            generator_model = config.get("generator_model", cls.DEFAULTS["generator_model"])
+            local_model_path = f"{config['local_models_path']}/Xenova/{generator_model}"
+            zip_path = f"{config['local_models_path']}/xenova-{generator_model}.zip"
+
+            # Check for zip file first, then unzip if needed
+            if os.path.exists(zip_path) and not os.path.exists(local_model_path):
+                print(f"Found zipped model at {zip_path}, extracting...")
+                import zipfile
+
+                with zipfile.ZipFile(zip_path, "r") as zipf:
+                    zipf.extractall(config["local_models_path"])
+                print(f"Extracted to {config['local_models_path']}")
+
+            if os.path.exists(local_model_path):
+                print(f"Found local model at {local_model_path}, using local files")
+                config["use_local_models"] = True
+                # Convert filesystem path to emfs: URI for browser access
+                # emfs: expects path relative to /drive, so models/Xenova/gpt2
+                # Remove /drive/ prefix to get models/Xenova/gpt2
+                emfs_path = local_model_path.replace("/drive/", "")
+                config["emfs_model_uri"] = f"emfs:{emfs_path}"
+                print(f"Using emfs URI: {config['emfs_model_uri']}")
+            else:
+                print(f"No local model found at {local_model_path}, will download from HuggingFace")
+
+        # Convert Python dict to JS object for compatibility with xeus-python
+        try:
+            # Try pyodide first (pyodide kernel)
+            from pyodide.ffi import to_js
+
+            js_config = to_js(config, dict_converter=lambda d: d)
+        except ImportError:
+            # In xeus-python, use pjs module (no dict_converter arg)
+            try:
+                from pjs import to_js
+
+                js_config = to_js(config)
+            except ImportError:
+                # Fallback: pass dict directly (works in some environments)
+                js_config = config
+        except TypeError:
+            # pyodide.ffi.to_js exists but dict_converter not supported
+            # Try simple to_js without dict_converter
+            try:
+                js_config = to_js(config)
+            except Exception:
+                js_config = config
+
+        instance = createLLM(js_config)
+        await instance.init()
+
+        return cls(instance)
+
+    # --- core API ---
+
+    async def generate(self, prompt: str):
+        return await self.js.generate(prompt)
+
+    async def embed(self, text: str):
+        vec = await self.js.embed(text)
+        return list(vec)
+
+    # --- similarity ---
+
+    async def similarity_search(self, query, docs):
+        from js import cosineSimilarityBatch
+
+        query_vec = await self.embed(query)
+
+        doc_vecs = []
+        for doc in docs:
+            doc_vecs.append(await self.embed(doc))
+
+        # Convert Python lists to JS arrays for the JS function
+        try:
+            from pyodide.ffi import to_js
+
+            js_query_vec = to_js(query_vec)
+            js_doc_vecs = to_js(doc_vecs)
+        except ImportError:
+            try:
+                from pjs import to_js
+
+                js_query_vec = to_js(query_vec)
+                js_doc_vecs = to_js(doc_vecs)
+            except ImportError:
+                # Fallback: hope it works
+                js_query_vec = query_vec
+                js_doc_vecs = doc_vecs
+
+        sims = cosineSimilarityBatch(js_query_vec, js_doc_vecs)
+
+        results = list(zip(docs, sims))
+        results.sort(key=lambda x: x[1], reverse=True)
+
+        return results
+
+    # --- model export ---
+
+    async def export_model_files(self, model_name=None, as_zip=True):
+        """Export model files from browser cache to /drive for download.
+
+        Args:
+            model_name: Name of the model to export (default: current generator model)
+            as_zip: If True, export as a zip file; if False, export as individual files
+        """
+        import os
+        import zipfile
+
+        if model_name is None:
+            model_name = self.DEFAULTS["generator_model"]
+
+        # Check if model is already exported
+        if as_zip:
+            zip_path = f"/drive/models/xenova-{model_name}.zip"
+            if os.path.exists(zip_path):
+                message = f"Model already exported at {zip_path}"
+                print(message)
+                return message
+        else:
+            model_dir = f"/drive/models/Xenova/{model_name}"
+            if os.path.exists(model_dir) and os.listdir(model_dir):
+                message = f"Model already exported at {model_dir}"
+                print(message)
+                return message
+
+        result = await self.js.exportModelFiles(model_name)
+
+        # Check if result is an object with files (Python-side export needed)
+        if hasattr(result, "success") and result.success:
+            print(f"Exporting {len(result.files)} files for {result.model}")
+
+            if as_zip:
+                # Export as zip file
+                zip_path = f"/drive/models/xenova-{result.model}.zip"
+                os.makedirs("/drive/models", exist_ok=True)
+
+                # Deduplicate files by name, keeping the largest version
+                files_by_name = {}
+                for file_obj in result.files:
+                    filename = file_obj.filename
+                    data = file_obj.data
+                    file_bytes = bytes(data.to_py())
+
+                    if filename not in files_by_name or len(file_bytes) > len(files_by_name[filename]):
+                        files_by_name[filename] = file_bytes
+
+                with zipfile.ZipFile(zip_path, "w", zipfile.ZIP_DEFLATED) as zipf:
+                    for filename, file_bytes in files_by_name.items():
+                        # Store with Xenova/{model}/ prefix
+                        arcname = f"Xenova/{result.model}/{filename}"
+                        zipf.writestr(arcname, file_bytes)
+                        print(f"Added to zip: {arcname} ({len(file_bytes)} bytes)")
+
+                message = f"Successfully exported {len(files_by_name)} files to {zip_path}"
+                print(message)
+                return message
+            else:
+                # Export as individual files
+                model_dir = f"/drive/models/Xenova/{result.model}"
+                os.makedirs(model_dir, exist_ok=True)
+                print(f"Created directory: {model_dir}")
+
+                # Deduplicate files by name, keeping the largest version
+                files_by_name = {}
+                for file_obj in result.files:
+                    filename = file_obj.filename
+                    data = file_obj.data
+                    file_bytes = bytes(data.to_py())
+
+                    if filename not in files_by_name or len(file_bytes) > len(files_by_name[filename]):
+                        files_by_name[filename] = file_bytes
+
+                for filename, file_bytes in files_by_name.items():
+                    filepath = f"{model_dir}/{filename}"
+                    with open(filepath, "wb") as f:
+                        f.write(file_bytes)
+
+                    print(f"Exported: {filepath} ({len(file_bytes)} bytes)")
+
+                message = f"Successfully exported {len(files_by_name)} files to {model_dir}"
+                print(message)
+                return message
+        else:
+            # String result from JS side
+            print(f"Export result: {result}")
+            return result

lmlite/version.py

@@ -0,0 +1,4 @@
+# Copyright 2026 Emmanuel Decitre
+# SPDX-License-Identifier: Apache-2.0
+
+__version__ = "0.1.0"

lmlite-0.2.1.dist-info/METADATA

@@ -0,0 +1,272 @@
+Metadata-Version: 2.4
+Name: lmlite
+Version: 0.2.1
+Summary: Browser-native LLM orchestration in JupyterLite
+Author-email: Emmanuel Decitre <590094+decitre@users.noreply.github.com>
+License-Expression: Apache-2.0
+Project-URL: repository, https://github.com/decitre/lmlite
+Classifier: Development Status :: 4 - Beta
+Classifier: Environment :: WebAssembly :: Emscripten
+Classifier: Intended Audience :: Developers
+Classifier: Natural Language :: English
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Programming Language :: Python :: 3 :: Only
+Classifier: Programming Language :: Python :: Implementation :: CPython
+Requires-Python: >=3.11
+Description-Content-Type: text/markdown
+License-File: LICENSE
+License-File: NOTICE
+Dynamic: license-file
+
+# lmLite
+
+Browser-native LLM orchestration in JupyterLite - run language models entirely in your browser using WebAssembly.
+
+## Features
+
+- **Browser-native**: Run LLMs entirely in the browser, no server required
+- **Zero setup**: No installation, no API keys, works offline
+- **Privacy-first**: All computation happens locally in your browser
+- **Multiple models**: Support for GPT-2, DistilGPT-2, and more via Transformers.js
+- **Embeddings**: Generate text embeddings for semantic search
+- **Similarity search**: Built-in cosine similarity for RAG applications
+- **Model caching**: Download once, use offline forever
+- **Pythonic API**: Clean, async Python interface
+
+<!-- To re-introduce when the Actions page is up
+## Quick Start
+
+### Try the Demo (No Installation)
+
+**Recommended**: Visit the live demo with xeus-python kernel at:
+[https://decitre.github.io/lmlite/](https://decitre.github.io/lmlite/)
+
+**Alternative**: Pyodide kernel demo at:
+[https://decitre.github.io/lmlite/pyodide/](https://decitre.github.io/lmlite/pyodide/)
+
+Open the `demo.ipynb` notebook and run the cells to see LMLite in action.
+
+### Install from PyPI
+
+```bash
+pip install lmlite
+```
+
+**Note**: LMLite is designed to run in JupyterLite environments (xeus-python or pyodide kernels). For local development and testing, see the Development section below.
+
+-->
+
+## Usage
+
+```python
+from lmlite import LLM
+
+# Create LLM instance (downloads model on first run)
+llm = await LLM.create(generator_model="gpt2")
+
+# Generate text
+text = await llm.generate("Python is a great language because")
+print(text)
+
+# Generate embeddings
+embedding = await llm.embed("Hello world")
+print(embedding[:5])  # First 5 dimensions
+
+# Similarity search
+docs = [
+    "JupyterLite runs entirely in the browser.",
+    "Python is widely used for machine learning.",
+    "TypeScript is great for frontend applications.",
+]
+
+results = await llm.similarity_search(
+    "Where does JupyterLite run?",
+    docs
+)
+
+for doc, score in results:
+    print(f"{score:.3f} -> {doc}")
+```
+
+### Configuration Options
+
+```python
+llm = await LLM.create(
+    generator_model="gpt2",           # or "distilgpt2", etc.
+    embedding_model="all-MiniLM-L6-v2",
+    max_new_tokens=50,
+    temperature=0.7,
+    top_k=50,
+    do_sample=True,
+    use_local_models=False,           # Auto-detect local models
+    local_models_path="/drive/models"
+)
+```
+
+### Export Models for Offline Use
+
+After using models in the browser, you can export them for offline use:
+
+```python
+# Export to zip file (default)
+await llm.export_model_files("gpt2")
+
+# Export to directory
+await llm.export_model_files("gpt2", as_zip=False)
+```
+
+The exported files will be saved to `/drive/models/` and can be downloaded from JupyterLite.
+
+## Development
+
+### Prerequisites
+
+- [pixi](https://pixi.sh/) - Package manager (required)
+- [micromamba](https://mamba.readthedocs.io/en/latest/installation/micromamba-installation.html) - Conda package manager (required for JupyterLite builds)
+
+**Supported Platforms**: macOS (Intel/Apple Silicon) and Linux. Windows is not currently supported for development.
+
+### Setup
+
+```bash
+# Clone the repository
+git clone https://github.com/decitre/lmlite.git
+cd lmlite
+pixi install
+```
+
+### Development Tasks
+
+```bash
+# Run tests
+pixi run test
+
+# Quick tests (skip notebook tests)
+pixi run quick-test
+
+# Run linter
+pixi run lint
+
+# Check linting without fixing
+pixi run lint-check
+
+# Run tests with coverage
+pixi run coverage
+
+# Build wheel
+pixi run wheel
+```
+
+### Build JupyterLite Demo
+
+```bash
+pixi run lite-build
+```
+
+Follow the instructions provided by the command.
+
+### Run Tests in Different Python Versions
+
+```bash
+pixi run --environment py311 test
+pixi run --environment py312 test
+pixi run --environment py313 test
+```
+
+## How It Works
+
+LMLite bridges Python (via xeus-python or pyodide) and JavaScript (via Transformers.js):
+
+1. **JavaScript Layer**: Uses [@huggingface/transformers](https://www.npmjs.com/package/@xenova/transformers) to run ONNX models in the browser
+2. **Python Bridge**: Exposes JavaScript functionality through a Pythonic async API
+3. **Kernel Support**:
+   - **xeus-python** (recommended): Full CPython in WebAssembly via emscripten
+   - **pyodide**: Alternative WebAssembly Python runtime
+4. **Model Loading**:
+   - First run: Downloads models from HuggingFace CDN
+   - Cached: Uses browser's Cache API or IndexedDB
+   - Local: Reads from virtual filesystem if available
+5. **Execution**: Models run entirely in-browser using WebAssembly (WASM)
+
+## Supported Models
+
+### Text Generation
+- `gpt2` - GPT-2 (124M parameters)
+- `distilgpt2` - Smaller, faster GPT-2 variant
+
+### Embeddings
+- `all-MiniLM-L6-v2` - Sentence embeddings (384 dimensions)
+
+For other models, check [Xenova's model list](https://huggingface.co/Xenova).
+
+## Architecture
+
+```
+┌─────────────────────────────────────┐
+│   Python (JupyterLite)              │
+│   ├─ xeus-python (recommended)      │
+│   └─ pyodide (alternative)          │
+│                                     │
+│   from lmlite import LLM            │
+│   llm = await LLM.create()          │
+│   text = await llm.generate(...)    │
+└─────────────┬───────────────────────┘
+              │ Bridge
+              │ (pyodide.ffi / pjs)
+┌─────────────▼───────────────────────┐
+│   JavaScript (Browser)              │
+│                                     │
+│   Transformers.js                   │
+│   ├─ Model loading                  │
+│   ├─ ONNX Runtime (WASM)            │
+│   └─ WebGPU (optional)              │
+└─────────────────────────────────────┘
+```
+
+
+## License
+
+This project is licensed under the Apache License 2.0 - see the LICENSE file for details.
+
+**Dependencies:**
+- [Transformers.js](https://github.com/xenova/transformers.js) - Apache 2.0
+
+**Note**: Individual model licenses may vary. Check the model card on HuggingFace before use.
+
+## Acknowledgments
+
+- [Transformers.js](https://github.com/xenova/transformers.js) by [@xenova](https://github.com/xenova)
+- [JupyterLite](https://jupyterlite.readthedocs.io/) team
+- [Pyodide](https://pyodide.org/) project
+- [Hugging Face](https://huggingface.co/) for model hosting
+- [LLMs running in the browser](https://thekevinscott.com/llms-in-the-browser/)
+
+## Troubleshooting
+
+### Models downloading every time?
+
+Check browser console for Cache API availability. Some privacy settings may disable caching.
+
+### Out of memory errors?
+
+Try smaller models like `distilgpt2` or reduce `max_new_tokens`.
+
+### CORS errors?
+
+Ensure you're running from `http://localhost` or a proper HTTPS domain, not `file://`.
+
+## Citation
+
+If you use LMLite in your research, please cite:
+
+```bibtex
+@software{lmlite2026,
+  author = {Decitre, Emmanuel},
+  title = {LMLite: Browser-native LLM orchestration in JupyterLite},
+  year = {2026},
+  url = {https://github.com/decitre/lmlite}
+}
+```

lmlite-0.2.1.dist-info/RECORD

@@ -0,0 +1,11 @@
+lmlite/__init__.py,sha256=XKSd_auljtgczvh7YNB4vqVYqFqZbNSmlOUnwWvOCOk,151
+lmlite/llm.py,sha256=1CVJIF4rO2NcwQsNKdveUi6Gy7afEcR933gULkSYCkw,8408
+lmlite/version.py,sha256=v-6nCiQAt-cqjyvyREVTlAC42XTscA_w0tiWu9yJldU,95
+lmlite/bridge/runtime.py,sha256=_w2PTLv6mgV0ycYjzExUaIVb9HMcakw7cXHlepy-SHw,449
+lmlite/js/llm.js,sha256=d6O52LXCeSdN_ehLNsGfSYx63c44UBp1HSjORCC1ls0,15938
+lmlite-0.2.1.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+lmlite-0.2.1.dist-info/licenses/NOTICE,sha256=jgPBqmkAo8vp52qAAlCbpgoXqsFAKlxSXesIVFwcwQY,176
+lmlite-0.2.1.dist-info/METADATA,sha256=YlOjR6f9QnxH1GzptVM_PAedAZ9jf5J0Ftj9ydHzJ4g,8126
+lmlite-0.2.1.dist-info/WHEEL,sha256=aeYiig01lYGDzBgS8HxWXOg3uV61G9ijOsup-k9o1sk,91
+lmlite-0.2.1.dist-info/top_level.txt,sha256=TAp9F2hPJrEtKLWwbtY_1AUiJOGbV8YvKu2BfJe3Ra8,7
+lmlite-0.2.1.dist-info/RECORD,,

lmlite-0.2.1.dist-info/WHEEL

@@ -0,0 +1,5 @@
+Wheel-Version: 1.0
+Generator: setuptools (82.0.1)
+Root-Is-Purelib: true
+Tag: py3-none-any
+

lmlite-0.2.1.dist-info/licenses/LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

lmlite-0.2.1.dist-info/licenses/NOTICE

@@ -0,0 +1,6 @@
+lmLite
+
+This project includes code developed by Emmanuel Decitre.
+
+It uses Transformers.js (Apache License 2.0).
+Models loaded at runtime may be subject to their own licenses.

lmlite-0.2.1.dist-info/top_level.txt

@@ -0,0 +1 @@
+lmlite