@stacksjs/ai 0.70.53 → 0.70.55

package/src/drivers/ollama/index.ts

@@ -0,0 +1,514 @@
+/**
+ * Ollama API Driver
+ *
+ * Local LLM integration via Ollama.
+ * Supports chat completions, streaming, and embeddings.
+ */
+
+import type { AIDriver, AIDriverConfig, AIMessage, AIResult, ChatCompletionOptions, OllamaAPIResponse } from '../../types'
+
+export interface OllamaDriverConfig extends AIDriverConfig {
+  host?: string
+  model?: string
+  embeddingModel?: string
+}
+
+const DEFAULT_HOST = 'http://localhost:11434'
+const DEFAULT_MODEL = 'llama3.2'
+const DEFAULT_EMBEDDING_MODEL = 'nomic-embed-text'
+
+let globalConfig: OllamaDriverConfig = {}
+
+/**
+ * Configure Ollama globally
+ */
+export function configure(config: OllamaDriverConfig): void {
+  globalConfig = { ...globalConfig, ...config }
+}
+
+function getConfig(config?: Partial<OllamaDriverConfig>): OllamaDriverConfig {
+  return {
+    host: config?.host || globalConfig.host || process.env.OLLAMA_HOST || DEFAULT_HOST,
+    model: config?.model || globalConfig.model || process.env.OLLAMA_MODEL || DEFAULT_MODEL,
+    embeddingModel: config?.embeddingModel || globalConfig.embeddingModel || DEFAULT_EMBEDDING_MODEL,
+  }
+}
+
+export function createOllamaDriver(config: OllamaDriverConfig = {}): AIDriver {
+  const {
+    host = process.env.OLLAMA_HOST || DEFAULT_HOST,
+    model = process.env.OLLAMA_MODEL || DEFAULT_MODEL,
+    embeddingModel = DEFAULT_EMBEDDING_MODEL,
+  } = config
+
+  return {
+    name: 'Ollama',
+
+    async process(command: string, systemPrompt: string, history: AIMessage[]): Promise<string> {
+      const response = await fetch(`${host}/api/chat`, {
+        method: 'POST',
+        headers: { 'Content-Type': 'application/json' },
+        body: JSON.stringify({
+          model,
+          messages: [
+            { role: 'system', content: systemPrompt },
+            ...history,
+            { role: 'user', content: command },
+          ],
+          stream: false,
+        }),
+      })
+
+      if (!response.ok) {
+        const error = await response.text()
+        throw new Error(`Ollama API error: ${error}`)
+      }
+
+      const data = (await response.json()) as OllamaAPIResponse
+      return data.message.content
+    },
+
+    async *stream(command: string, systemPrompt: string, history: AIMessage[]): AsyncGenerator<string> {
+      const response = await fetch(`${host}/api/chat`, {
+        method: 'POST',
+        headers: { 'Content-Type': 'application/json' },
+        body: JSON.stringify({
+          model,
+          messages: [
+            { role: 'system', content: systemPrompt },
+            ...history,
+            { role: 'user', content: command },
+          ],
+          stream: true,
+        }),
+      })
+
+      if (!response.ok) {
+        const error = await response.text()
+        throw new Error(`Ollama API error: ${error}`)
+      }
+
+      const reader = response.body?.getReader()
+      if (!reader) throw new Error('No response body')
+
+      const decoder = new TextDecoder()
+      let buffer = ''
+
+      while (true) {
+        const { done, value } = await reader.read()
+        if (done) break
+
+        buffer += decoder.decode(value, { stream: true })
+        const lines = buffer.split('\n')
+        buffer = lines.pop() || ''
+
+        for (const line of lines) {
+          if (!line.trim()) continue
+
+          try {
+            const data = JSON.parse(line) as OllamaAPIResponse
+            if (data.message?.content) {
+              yield data.message.content
+            }
+          }
+          catch {
+            // Skip invalid JSON
+          }
+        }
+      }
+    },
+
+    async embed(input: string | string[]): Promise<number[] | number[][]> {
+      const inputs = Array.isArray(input) ? input : [input]
+      const embeddings: number[][] = []
+
+      for (const text of inputs) {
+        const response = await fetch(`${host}/api/embeddings`, {
+          method: 'POST',
+          headers: { 'Content-Type': 'application/json' },
+          body: JSON.stringify({
+            model: embeddingModel,
+            prompt: text,
+          }),
+        })
+
+        if (!response.ok) {
+          const error = await response.text()
+          throw new Error(`Ollama Embeddings API error: ${error}`)
+        }
+
+        const data = (await response.json()) as { embedding: number[] }
+        embeddings.push(data.embedding)
+      }
+
+      return Array.isArray(input) ? embeddings : embeddings[0]
+    },
+  }
+}
+
+/**
+ * Chat completion with full options
+ */
+export async function chat(
+  messages: AIMessage[],
+  options: ChatCompletionOptions = {},
+): Promise<AIResult> {
+  const config = getConfig()
+  const {
+    model = config.model,
+    temperature,
+    topP,
+    stop,
+  } = options
+
+  const response = await fetch(`${config.host}/api/chat`, {
+    method: 'POST',
+    headers: { 'Content-Type': 'application/json' },
+    body: JSON.stringify({
+      model,
+      messages,
+      stream: false,
+      options: {
+        temperature,
+        top_p: topP,
+        stop,
+      },
+    }),
+  })
+
+  if (!response.ok) {
+    const error = await response.text()
+    throw new Error(`Ollama API error: ${error}`)
+  }
+
+  const data = (await response.json()) as any
+
+  return {
+    content: data.message.content,
+    model: data.model,
+    usage: {
+      promptTokens: data.prompt_eval_count || 0,
+      completionTokens: data.eval_count || 0,
+      totalTokens: (data.prompt_eval_count || 0) + (data.eval_count || 0),
+    },
+    finishReason: data.done_reason || 'stop',
+  }
+}
+
+/**
+ * Stream chat completion
+ */
+export async function* streamChat(
+  messages: AIMessage[],
+  options: ChatCompletionOptions = {},
+): AsyncGenerator<string> {
+  const config = getConfig()
+  const {
+    model = config.model,
+    temperature,
+    topP,
+    stop,
+  } = options
+
+  const response = await fetch(`${config.host}/api/chat`, {
+    method: 'POST',
+    headers: { 'Content-Type': 'application/json' },
+    body: JSON.stringify({
+      model,
+      messages,
+      stream: true,
+      options: {
+        temperature,
+        top_p: topP,
+        stop,
+      },
+    }),
+  })
+
+  if (!response.ok) {
+    const error = await response.text()
+    throw new Error(`Ollama API error: ${error}`)
+  }
+
+  const reader = response.body?.getReader()
+  if (!reader) throw new Error('No response body')
+
+  const decoder = new TextDecoder()
+  let buffer = ''
+
+  while (true) {
+    const { done, value } = await reader.read()
+    if (done) break
+
+    buffer += decoder.decode(value, { stream: true })
+    const lines = buffer.split('\n')
+    buffer = lines.pop() || ''
+
+    for (const line of lines) {
+      if (!line.trim()) continue
+
+      try {
+        const data = JSON.parse(line) as OllamaAPIResponse
+        if (data.message?.content) {
+          yield data.message.content
+        }
+      }
+      catch {
+        // Skip invalid JSON
+      }
+    }
+  }
+}
+
+/**
+ * Generate text completion (non-chat)
+ */
+export async function generate(
+  prompt: string,
+  options: {
+    model?: string
+    system?: string
+    template?: string
+    context?: number[]
+    raw?: boolean
+    format?: 'json'
+    images?: string[]
+  } = {},
+): Promise<AIResult> {
+  const config = getConfig()
+
+  const response = await fetch(`${config.host}/api/generate`, {
+    method: 'POST',
+    headers: { 'Content-Type': 'application/json' },
+    body: JSON.stringify({
+      model: options.model || config.model,
+      prompt,
+      system: options.system,
+      template: options.template,
+      context: options.context,
+      raw: options.raw,
+      format: options.format,
+      images: options.images,
+      stream: false,
+    }),
+  })
+
+  if (!response.ok) {
+    const error = await response.text()
+    throw new Error(`Ollama API error: ${error}`)
+  }
+
+  const data = (await response.json()) as any
+
+  return {
+    content: data.response,
+    model: data.model,
+    usage: {
+      promptTokens: data.prompt_eval_count || 0,
+      completionTokens: data.eval_count || 0,
+      totalTokens: (data.prompt_eval_count || 0) + (data.eval_count || 0),
+    },
+  }
+}
+
+/**
+ * Create embeddings
+ */
+export async function embed(
+  input: string | string[],
+  model?: string,
+): Promise<number[] | number[][]> {
+  const config = getConfig()
+  const embeddingModel = model || config.embeddingModel
+
+  const inputs = Array.isArray(input) ? input : [input]
+  const embeddings: number[][] = []
+
+  for (const text of inputs) {
+    const response = await fetch(`${config.host}/api/embeddings`, {
+      method: 'POST',
+      headers: { 'Content-Type': 'application/json' },
+      body: JSON.stringify({
+        model: embeddingModel,
+        prompt: text,
+      }),
+    })
+
+    if (!response.ok) {
+      const error = await response.text()
+      throw new Error(`Ollama Embeddings API error: ${error}`)
+    }
+
+    const data = (await response.json()) as { embedding: number[] }
+    embeddings.push(data.embedding)
+  }
+
+  return Array.isArray(input) ? embeddings : embeddings[0]
+}
+
+/**
+ * List available models
+ */
+export async function listModels(): Promise<Array<{
+  name: string
+  modified_at: string
+  size: number
+  digest: string
+  details: {
+    format: string
+    family: string
+    families: string[]
+    parameter_size: string
+    quantization_level: string
+  }
+}>> {
+  const config = getConfig()
+
+  const response = await fetch(`${config.host}/api/tags`, {
+    method: 'GET',
+  })
+
+  if (!response.ok) {
+    const error = await response.text()
+    throw new Error(`Ollama API error: ${error}`)
+  }
+
+  const data = (await response.json()) as { models: any[] }
+  return data.models
+}
+
+/**
+ * Pull a model from the library
+ */
+export async function pullModel(
+  name: string,
+  onProgress?: (status: string, completed?: number, total?: number) => void,
+): Promise<void> {
+  const config = getConfig()
+
+  const response = await fetch(`${config.host}/api/pull`, {
+    method: 'POST',
+    headers: { 'Content-Type': 'application/json' },
+    body: JSON.stringify({ name, stream: true }),
+  })
+
+  if (!response.ok) {
+    const error = await response.text()
+    throw new Error(`Ollama API error: ${error}`)
+  }
+
+  const reader = response.body?.getReader()
+  if (!reader) return
+
+  const decoder = new TextDecoder()
+  let buffer = ''
+
+  while (true) {
+    const { done, value } = await reader.read()
+    if (done) break
+
+    buffer += decoder.decode(value, { stream: true })
+    const lines = buffer.split('\n')
+    buffer = lines.pop() || ''
+
+    for (const line of lines) {
+      if (!line.trim()) continue
+
+      try {
+        const data = JSON.parse(line) as { status?: string; completed?: number; total?: number }
+        if (onProgress) {
+          onProgress(data.status, data.completed, data.total)
+        }
+      }
+      catch {
+        // Skip invalid JSON
+      }
+    }
+  }
+}
+
+/**
+ * Delete a model
+ */
+export async function deleteModel(name: string): Promise<void> {
+  const config = getConfig()
+
+  const response = await fetch(`${config.host}/api/delete`, {
+    method: 'DELETE',
+    headers: { 'Content-Type': 'application/json' },
+    body: JSON.stringify({ name }),
+  })
+
+  if (!response.ok) {
+    const error = await response.text()
+    throw new Error(`Ollama API error: ${error}`)
+  }
+}
+
+/**
+ * Show model information
+ */
+export async function showModel(_name: string): Promise<{
+  modelfile: string
+  parameters: string
+  template: string
+  details: {
+    format: string
+    family: string
+    families: string[]
+    parameter_size: string
+    quantization_level: string
+  }
+}> {
+  const config = getConfig()
+
+  const response = await fetch(`${config.host}/api/show`, {
+    method: 'POST',
+    headers: { 'Content-Type': 'application/json' },
+    body: JSON.stringify({ name: _name }),
+  })
+
+  if (!response.ok) {
+    const error = await response.text()
+    throw new Error(`Ollama API error: ${error}`)
+  }
+
+  return response.json() as any
+}
+
+/**
+ * Check if Ollama is running
+ */
+export async function isRunning(): Promise<boolean> {
+  const config = getConfig()
+
+  try {
+    const response = await fetch(`${config.host}/api/tags`, {
+      method: 'GET',
+    })
+    return response.ok
+  }
+  catch {
+    return false
+  }
+}
+
+export const ollamaDriver: { create: typeof createOllamaDriver } = {
+  create: createOllamaDriver,
+}
+
+export const ollama = {
+  configure,
+  chat,
+  streamChat,
+  generate,
+  embed,
+  listModels,
+  pullModel,
+  deleteModel,
+  showModel,
+  isRunning,
+  createDriver: createOllamaDriver,
+}
+
+export default ollama