cognitive-modules-cli 1.4.0 → 2.2.0

package/src/providers/base.ts

@@ -1,17 +1,8 @@
 /**
  * Base Provider - Abstract class for all LLM providers
- * v2.5: Added streaming and multimodal support
  */
 
-import type { 
-  Provider, 
-  InvokeParams, 
-  InvokeResult,
-  ProviderV25,
-  InvokeParamsV25,
-  StreamingInvokeResult,
-  ModalityType
-} from '../types.js';
+import type { Provider, InvokeParams, InvokeResult } from '../types.js';
 
 export abstract class BaseProvider implements Provider {
   abstract name: string;
@@ -36,79 +27,3 @@ export abstract class BaseProvider implements Provider {
     }
   }
 }
-
-/**
- * Base Provider with v2.5 streaming and multimodal support
- */
-export abstract class BaseProviderV25 extends BaseProvider implements ProviderV25 {
-  /**
-   * Check if this provider supports streaming
-   * Override in subclass to enable streaming
-   */
-  supportsStreaming(): boolean {
-    return false;
-  }
-  
-  /**
-   * Check if this provider supports multimodal input/output
-   * Override in subclass to enable multimodal
-   */
-  supportsMultimodal(): { input: ModalityType[]; output: ModalityType[] } {
-    return {
-      input: ['text'],
-      output: ['text']
-    };
-  }
-  
-  /**
-   * Invoke with streaming response
-   * Override in subclass to implement streaming
-   */
-  async invokeStream(params: InvokeParamsV25): Promise<StreamingInvokeResult> {
-    // Default: fallback to non-streaming with async generator wrapper
-    const result = await this.invoke(params);
-    
-    async function* generateChunks(): AsyncIterable<string> {
-      yield result.content;
-    }
-    
-    return {
-      stream: generateChunks(),
-      usage: result.usage
-    };
-  }
-  
-  /**
-   * Format media inputs for the specific provider API
-   * Override in subclass for provider-specific formatting
-   */
-  protected formatMediaForProvider(
-    images?: Array<{ type: string; url?: string; data?: string; media_type?: string }>,
-    _audio?: Array<{ type: string; url?: string; data?: string; media_type?: string }>,
-    _video?: Array<{ type: string; url?: string; data?: string; media_type?: string }>
-  ): unknown[] {
-    // Default implementation for image-only providers (like OpenAI Vision)
-    if (!images || images.length === 0) {
-      return [];
-    }
-    
-    return images.map(img => {
-      if (img.type === 'url' && img.url) {
-        return {
-          type: 'image_url',
-          image_url: {
-            url: img.url
-          }
-        };
-      } else if (img.type === 'base64' && img.data && img.media_type) {
-        return {
-          type: 'image_url',
-          image_url: {
-            url: `data:${img.media_type};base64,${img.data}`
-          }
-        };
-      }
-      return null;
-    }).filter(Boolean);
-  }
-}

package/src/providers/openai.ts

@@ -1,35 +1,17 @@
 /**
  * OpenAI Provider - OpenAI API (and compatible APIs)
- * v2.5: Added streaming and multimodal (vision) support
  */
 
-import { BaseProviderV25 } from './base.js';
-import type { 
-  InvokeParams, 
-  InvokeResult, 
-  InvokeParamsV25, 
-  StreamingInvokeResult,
-  ModalityType,
-  MediaInput
-} from '../types.js';
+import { BaseProvider } from './base.js';
+import type { InvokeParams, InvokeResult } from '../types.js';
 
-// Type for OpenAI message content
-type OpenAIContentPart = 
-  | { type: 'text'; text: string }
-  | { type: 'image_url'; image_url: { url: string; detail?: 'low' | 'high' | 'auto' } };
-
-type OpenAIMessage = {
-  role: 'system' | 'user' | 'assistant';
-  content: string | OpenAIContentPart[];
-};
-
-export class OpenAIProvider extends BaseProviderV25 {
+export class OpenAIProvider extends BaseProvider {
   name = 'openai';
   private apiKey: string;
   private model: string;
   private baseUrl: string;
 
-  constructor(apiKey?: string, model = 'gpt-4o', baseUrl = 'https://api.openai.com/v1') {
+  constructor(apiKey?: string, model = 'gpt-5.2', baseUrl = 'https://api.openai.com/v1') {
     super();
     this.apiKey = apiKey || process.env.OPENAI_API_KEY || '';
     this.model = model;
@@ -40,27 +22,6 @@ export class OpenAIProvider extends BaseProviderV25 {
     return !!this.apiKey;
   }
 
-  /**
-   * Check if streaming is supported (always true for OpenAI)
-   */
-  supportsStreaming(): boolean {
-    return true;
-  }
-
-  /**
-   * Check multimodal support (vision models)
-   */
-  supportsMultimodal(): { input: ModalityType[]; output: ModalityType[] } {
-    // Vision models support image input
-    const visionModels = ['gpt-4o', 'gpt-4-vision', 'gpt-4-turbo', 'gpt-4o-mini'];
-    const supportsVision = visionModels.some(m => this.model.includes(m));
-    
-    return {
-      input: supportsVision ? ['text', 'image'] : ['text'],
-      output: ['text'] // DALL-E would be separate
-    };
-  }
-
   async invoke(params: InvokeParams): Promise<InvokeResult> {
     if (!this.isConfigured()) {
       throw new Error('OpenAI API key not configured. Set OPENAI_API_KEY environment variable.');
@@ -120,187 +81,4 @@ export class OpenAIProvider extends BaseProviderV25 {
       } : undefined,
     };
   }
-
-  /**
-   * Invoke with streaming response
-   */
-  async invokeStream(params: InvokeParamsV25): Promise<StreamingInvokeResult> {
-    if (!this.isConfigured()) {
-      throw new Error('OpenAI API key not configured. Set OPENAI_API_KEY environment variable.');
-    }
-
-    const url = `${this.baseUrl}/chat/completions`;
-
-    // Build messages with multimodal content if present
-    const messages = this.buildMessagesWithMedia(params);
-
-    const body: Record<string, unknown> = {
-      model: this.model,
-      messages,
-      temperature: params.temperature ?? 0.7,
-      max_tokens: params.maxTokens ?? 4096,
-      stream: true,
-    };
-
-    // Add JSON mode if schema provided
-    if (params.jsonSchema) {
-      body.response_format = { type: 'json_object' };
-    }
-
-    const response = await fetch(url, {
-      method: 'POST',
-      headers: {
-        'Content-Type': 'application/json',
-        'Authorization': `Bearer ${this.apiKey}`,
-      },
-      body: JSON.stringify(body),
-    });
-
-    if (!response.ok) {
-      const error = await response.text();
-      throw new Error(`OpenAI API error: ${response.status} - ${error}`);
-    }
-
-    const bodyReader = response.body?.getReader();
-    if (!bodyReader) {
-      throw new Error('No response body');
-    }
-
-    const decoder = new TextDecoder();
-    let usage: { promptTokens: number; completionTokens: number; totalTokens: number } | undefined;
-
-    // Capture reader reference for closure
-    const reader = bodyReader;
-
-    // Create async generator for streaming
-    async function* streamGenerator(): AsyncIterable<string> {
-      let buffer = '';
-      
-      while (true) {
-        const { done, value } = await reader.read();
-        
-        if (done) break;
-        
-        buffer += decoder.decode(value, { stream: true });
-        
-        // Parse SSE events
-        const lines = buffer.split('\n');
-        buffer = lines.pop() || '';
-        
-        for (const line of lines) {
-          if (line.startsWith('data: ')) {
-            const data = line.slice(6);
-            
-            if (data === '[DONE]') {
-              return;
-            }
-            
-            try {
-              const parsed = JSON.parse(data) as {
-                choices?: Array<{ delta?: { content?: string } }>;
-                usage?: { prompt_tokens?: number; completion_tokens?: number; total_tokens?: number };
-              };
-              
-              const content = parsed.choices?.[0]?.delta?.content;
-              if (content) {
-                yield content;
-              }
-              
-              // Capture usage if available
-              if (parsed.usage) {
-                usage = {
-                  promptTokens: parsed.usage.prompt_tokens || 0,
-                  completionTokens: parsed.usage.completion_tokens || 0,
-                  totalTokens: parsed.usage.total_tokens || 0,
-                };
-              }
-            } catch {
-              // Skip malformed JSON
-            }
-          }
-        }
-      }
-    }
-
-    return {
-      stream: streamGenerator(),
-      usage
-    };
-  }
-
-  /**
-   * Build messages with multimodal content (images)
-   */
-  private buildMessagesWithMedia(params: InvokeParamsV25): OpenAIMessage[] {
-    const hasImages = params.images && params.images.length > 0;
-    
-    if (!hasImages) {
-      return params.messages;
-    }
-    
-    // Find the last user message and add images to it
-    const messages: OpenAIMessage[] = [];
-    const lastUserIdx = params.messages.findLastIndex(m => m.role === 'user');
-    
-    for (let i = 0; i < params.messages.length; i++) {
-      const msg = params.messages[i];
-      
-      if (i === lastUserIdx && hasImages) {
-        // Convert to multimodal content
-        const content: OpenAIContentPart[] = [
-          { type: 'text', text: msg.content }
-        ];
-        
-        // Add images
-        for (const img of params.images!) {
-          const imageUrl = this.mediaInputToUrl(img);
-          if (imageUrl) {
-            content.push({
-              type: 'image_url',
-              image_url: { url: imageUrl, detail: 'auto' }
-            });
-          }
-        }
-        
-        messages.push({ role: msg.role, content });
-      } else {
-        messages.push({ role: msg.role, content: msg.content });
-      }
-    }
-    
-    // Add JSON schema instruction if needed
-    if (params.jsonSchema && lastUserIdx >= 0) {
-      const lastMsg = messages[lastUserIdx];
-      if (typeof lastMsg.content === 'string') {
-        lastMsg.content = lastMsg.content + this.buildJsonPrompt(params.jsonSchema);
-      } else {
-        // Content is array, append to text part
-        const textPart = lastMsg.content.find(p => p.type === 'text');
-        if (textPart && textPart.type === 'text') {
-          textPart.text = textPart.text + this.buildJsonPrompt(params.jsonSchema);
-        }
-      }
-    }
-    
-    return messages;
-  }
-
-  /**
-   * Convert MediaInput to URL for OpenAI API
-   */
-  private mediaInputToUrl(media: MediaInput): string | null {
-    switch (media.type) {
-      case 'url':
-        return media.url;
-      case 'base64':
-        return `data:${media.media_type};base64,${media.data}`;
-      case 'file':
-        // File paths would need to be loaded first
-        // This should be handled by the runner before calling the provider
-        console.warn('[cognitive] File media input not pre-loaded, skipping');
-        return null;
-      default:
-        return null;
-    }
-  }
 }

package/src/types.ts

@@ -1,6 +1,6 @@
 /**
  * Cognitive Runtime - Core Types
- * Version 2.5 - With streaming response and multimodal support
+ * Version 2.2 - With Control/Data plane separation, tier, overflow, extensible enums
  */
 
 // =============================================================================
@@ -493,321 +493,3 @@ export function shouldEscalate<T>(
   
   return false;
 }
-
-// =============================================================================
-// v2.5 Streaming Types
-// =============================================================================
-
-/** Response mode configuration */
-export type ResponseMode = 'sync' | 'streaming' | 'both';
-
-/** Chunk type for streaming */
-export type ChunkType = 'delta' | 'snapshot';
-
-/** Response configuration in module.yaml */
-export interface ResponseConfig {
-  mode: ResponseMode;
-  chunk_type?: ChunkType;
-  buffer_size?: number;
-  heartbeat_interval_ms?: number;
-  max_duration_ms?: number;
-}
-
-/** Meta chunk - initial streaming response */
-export interface MetaChunk {
-  ok: true;
-  streaming: true;
-  session_id: string;
-  meta: Partial<EnvelopeMeta>;
-}
-
-/** Delta chunk - incremental content */
-export interface DeltaChunk {
-  chunk: {
-    seq: number;
-    type: 'delta';
-    field?: string;
-    delta: string;
-  };
-}
-
-/** Snapshot chunk - full state replacement */
-export interface SnapshotChunk {
-  chunk: {
-    seq: number;
-    type: 'snapshot';
-    field?: string;
-    data: unknown;
-  };
-}
-
-/** Progress chunk - progress update */
-export interface ProgressChunk {
-  progress: {
-    percent: number;
-    stage?: string;
-    message?: string;
-  };
-}
-
-/** Final chunk - completion signal */
-export interface FinalChunk {
-  final: true;
-  meta: EnvelopeMeta;
-  data: ModuleResultData;
-  usage?: {
-    input_tokens: number;
-    output_tokens: number;
-    total_tokens: number;
-  };
-}
-
-/** Error chunk during streaming */
-export interface ErrorChunk {
-  ok: false;
-  streaming: true;
-  session_id?: string;
-  error: {
-    code: string;
-    message: string;
-    recoverable?: boolean;
-  };
-  partial_data?: unknown;
-}
-
-/** Union of all streaming chunk types */
-export type StreamingChunk = 
-  | MetaChunk 
-  | DeltaChunk 
-  | SnapshotChunk 
-  | ProgressChunk 
-  | FinalChunk 
-  | ErrorChunk;
-
-/** Streaming session state */
-export interface StreamingSession {
-  session_id: string;
-  module_name: string;
-  started_at: number;
-  chunks_sent: number;
-  accumulated_data: Record<string, unknown>;
-  accumulated_text: Record<string, string>;
-}
-
-// =============================================================================
-// v2.5 Multimodal Types
-// =============================================================================
-
-/** Supported modality types */
-export type ModalityType = 'text' | 'image' | 'audio' | 'video' | 'document';
-
-/** Modalities configuration in module.yaml */
-export interface ModalitiesConfig {
-  input: ModalityType[];
-  output: ModalityType[];
-  constraints?: MediaConstraints;
-}
-
-/** Media size/duration constraints */
-export interface MediaConstraints {
-  max_image_size_mb?: number;
-  max_audio_size_mb?: number;
-  max_video_size_mb?: number;
-  max_audio_duration_s?: number;
-  max_video_duration_s?: number;
-  allowed_image_types?: string[];
-  allowed_audio_types?: string[];
-  allowed_video_types?: string[];
-}
-
-/** Media input - URL reference */
-export interface UrlMediaInput {
-  type: 'url';
-  url: string;
-  media_type?: string;
-}
-
-/** Media input - Base64 inline */
-export interface Base64MediaInput {
-  type: 'base64';
-  media_type: string;
-  data: string;
-}
-
-/** Media input - File path */
-export interface FileMediaInput {
-  type: 'file';
-  path: string;
-}
-
-/** Union of media input types */
-export type MediaInput = UrlMediaInput | Base64MediaInput | FileMediaInput;
-
-/** Media output with metadata */
-export interface MediaOutput {
-  type: 'url' | 'base64' | 'file';
-  media_type: string;
-  url?: string;
-  data?: string;
-  path?: string;
-  width?: number;
-  height?: number;
-  duration_ms?: number;
-  expires_at?: string;
-  generation_params?: Record<string, unknown>;
-}
-
-/** Supported image MIME types */
-export const SUPPORTED_IMAGE_TYPES = [
-  'image/jpeg',
-  'image/png',
-  'image/webp',
-  'image/gif'
-] as const;
-
-/** Supported audio MIME types */
-export const SUPPORTED_AUDIO_TYPES = [
-  'audio/mpeg',
-  'audio/wav',
-  'audio/ogg',
-  'audio/webm'
-] as const;
-
-/** Supported video MIME types */
-export const SUPPORTED_VIDEO_TYPES = [
-  'video/mp4',
-  'video/webm',
-  'video/quicktime'
-] as const;
-
-// =============================================================================
-// v2.5 Error Codes
-// =============================================================================
-
-/** v2.5 Error codes for streaming and multimodal */
-export const ErrorCodesV25 = {
-  // Media errors (E1xxx)
-  UNSUPPORTED_MEDIA_TYPE: 'E1010',
-  MEDIA_TOO_LARGE: 'E1011',
-  MEDIA_FETCH_FAILED: 'E1012',
-  MEDIA_DECODE_FAILED: 'E1013',
-  
-  // Streaming errors (E2xxx)
-  STREAM_INTERRUPTED: 'E2010',
-  STREAM_TIMEOUT: 'E2011',
-  
-  // Capability errors (E4xxx)
-  STREAMING_NOT_SUPPORTED: 'E4010',
-  MULTIMODAL_NOT_SUPPORTED: 'E4011',
-} as const;
-
-export type ErrorCodeV25 = typeof ErrorCodesV25[keyof typeof ErrorCodesV25];
-
-// =============================================================================
-// v2.5 Runtime Capabilities
-// =============================================================================
-
-/** Runtime capability declaration */
-export interface RuntimeCapabilities {
-  streaming: boolean;
-  multimodal: {
-    input: ModalityType[];
-    output: ModalityType[];
-  };
-  max_media_size_mb: number;
-  supported_transports: ('sse' | 'websocket' | 'ndjson')[];
-}
-
-/** Default runtime capabilities */
-export const DEFAULT_RUNTIME_CAPABILITIES: RuntimeCapabilities = {
-  streaming: true,
-  multimodal: {
-    input: ['text', 'image'],
-    output: ['text']
-  },
-  max_media_size_mb: 20,
-  supported_transports: ['sse', 'ndjson']
-};
-
-// =============================================================================
-// v2.5 Extended Provider Interface
-// =============================================================================
-
-/** Extended invoke params with streaming support */
-export interface InvokeParamsV25 extends InvokeParams {
-  stream?: boolean;
-  images?: MediaInput[];
-  audio?: MediaInput[];
-  video?: MediaInput[];
-}
-
-/** Streaming invoke result */
-export interface StreamingInvokeResult {
-  stream: AsyncIterable<string>;
-  usage?: {
-    promptTokens: number;
-    completionTokens: number;
-    totalTokens: number;
-  };
-}
-
-/** Extended provider interface for v2.5 */
-export interface ProviderV25 extends Provider {
-  /** Check if provider supports streaming */
-  supportsStreaming?(): boolean;
-  
-  /** Check if provider supports multimodal input */
-  supportsMultimodal?(): { input: ModalityType[]; output: ModalityType[] };
-  
-  /** Invoke with streaming */
-  invokeStream?(params: InvokeParamsV25): Promise<StreamingInvokeResult>;
-}
-
-/** Type guard for v2.5 provider */
-export function isProviderV25(provider: Provider): provider is ProviderV25 {
-  return 'invokeStream' in provider || 'supportsStreaming' in provider;
-}
-
-// =============================================================================
-// v2.5 Module Configuration Extensions
-// =============================================================================
-
-/** Extended module interface for v2.5 */
-export interface CognitiveModuleV25 extends CognitiveModule {
-  /** v2.5: Response configuration */
-  response?: ResponseConfig;
-  
-  /** v2.5: Modalities configuration */
-  modalities?: ModalitiesConfig;
-}
-
-/** Type guard for v2.5 module */
-export function isModuleV25(module: CognitiveModule): module is CognitiveModuleV25 {
-  return 'response' in module || 'modalities' in module;
-}
-
-/** Check if module supports streaming */
-export function moduleSupportsStreaming(module: CognitiveModule): boolean {
-  if (!isModuleV25(module)) return false;
-  const mode = module.response?.mode;
-  return mode === 'streaming' || mode === 'both';
-}
-
-/** Check if module supports multimodal input */
-export function moduleSupportsMultimodal(module: CognitiveModule): boolean {
-  if (!isModuleV25(module)) return false;
-  const modalities = module.modalities?.input ?? ['text'];
-  return modalities.some(m => m !== 'text');
-}
-
-/** Get supported input modalities for module */
-export function getModuleInputModalities(module: CognitiveModule): ModalityType[] {
-  if (!isModuleV25(module)) return ['text'];
-  return module.modalities?.input ?? ['text'];
-}
-
-/** Get supported output modalities for module */
-export function getModuleOutputModalities(module: CognitiveModule): ModalityType[] {
-  if (!isModuleV25(module)) return ['text'];
-  return module.modalities?.output ?? ['text'];
-}