gencode-ai 0.1.2 → 0.2.0

package/src/pricing/models.ts

@@ -0,0 +1,334 @@
+/**
+ * Model pricing database
+ * Prices are per 1M tokens in USD
+ * Updated: January 2025
+ */
+
+import { ModelPricing } from './types.js';
+
+/**
+ * Pricing data for all supported models
+ * Source: Official provider pricing pages as of January 2025
+ */
+export const MODEL_PRICING: ModelPricing[] = [
+  // Anthropic Claude Models
+  {
+    provider: 'anthropic',
+    model: 'claude-opus-4-5',
+    inputPer1M: 15.0,
+    outputPer1M: 75.0,
+    effectiveDate: '2025-01-01',
+  },
+  {
+    provider: 'anthropic',
+    model: 'claude-opus-4',
+    inputPer1M: 15.0,
+    outputPer1M: 75.0,
+    effectiveDate: '2024-11-01',
+  },
+  {
+    provider: 'anthropic',
+    model: 'claude-sonnet-4-5',
+    inputPer1M: 3.0,
+    outputPer1M: 15.0,
+    effectiveDate: '2025-01-01',
+  },
+  {
+    provider: 'anthropic',
+    model: 'claude-sonnet-4',
+    inputPer1M: 3.0,
+    outputPer1M: 15.0,
+    effectiveDate: '2024-10-22',
+  },
+  {
+    provider: 'anthropic',
+    model: 'claude-3-5-sonnet-20241022',
+    inputPer1M: 3.0,
+    outputPer1M: 15.0,
+    effectiveDate: '2024-10-22',
+  },
+  {
+    provider: 'anthropic',
+    model: 'claude-3-5-sonnet-20240620',
+    inputPer1M: 3.0,
+    outputPer1M: 15.0,
+    effectiveDate: '2024-06-20',
+  },
+  {
+    provider: 'anthropic',
+    model: 'claude-haiku-3-5',
+    inputPer1M: 0.8,
+    outputPer1M: 4.0,
+    effectiveDate: '2024-11-01',
+  },
+  {
+    provider: 'anthropic',
+    model: 'claude-3-5-haiku-20241022',
+    inputPer1M: 0.8,
+    outputPer1M: 4.0,
+    effectiveDate: '2024-10-22',
+  },
+  {
+    provider: 'anthropic',
+    model: 'claude-3-haiku-20240307',
+    inputPer1M: 0.25,
+    outputPer1M: 1.25,
+    effectiveDate: '2024-03-07',
+  },
+  {
+    provider: 'anthropic',
+    model: 'claude-3-opus-20240229',
+    inputPer1M: 15.0,
+    outputPer1M: 75.0,
+    effectiveDate: '2024-02-29',
+  },
+
+  // OpenAI Models
+  {
+    provider: 'openai',
+    model: 'gpt-4o',
+    inputPer1M: 2.5,
+    outputPer1M: 10.0,
+    effectiveDate: '2024-08-06',
+  },
+  {
+    provider: 'openai',
+    model: 'gpt-4o-2024-11-20',
+    inputPer1M: 2.5,
+    outputPer1M: 10.0,
+    effectiveDate: '2024-11-20',
+  },
+  {
+    provider: 'openai',
+    model: 'gpt-4o-2024-08-06',
+    inputPer1M: 2.5,
+    outputPer1M: 10.0,
+    effectiveDate: '2024-08-06',
+  },
+  {
+    provider: 'openai',
+    model: 'gpt-4o-2024-05-13',
+    inputPer1M: 5.0,
+    outputPer1M: 15.0,
+    effectiveDate: '2024-05-13',
+  },
+  {
+    provider: 'openai',
+    model: 'gpt-4o-mini',
+    inputPer1M: 0.15,
+    outputPer1M: 0.6,
+    effectiveDate: '2024-07-18',
+  },
+  {
+    provider: 'openai',
+    model: 'gpt-4o-mini-2024-07-18',
+    inputPer1M: 0.15,
+    outputPer1M: 0.6,
+    effectiveDate: '2024-07-18',
+  },
+  {
+    provider: 'openai',
+    model: 'gpt-4-turbo',
+    inputPer1M: 10.0,
+    outputPer1M: 30.0,
+    effectiveDate: '2024-04-09',
+  },
+  {
+    provider: 'openai',
+    model: 'gpt-4-turbo-2024-04-09',
+    inputPer1M: 10.0,
+    outputPer1M: 30.0,
+    effectiveDate: '2024-04-09',
+  },
+  {
+    provider: 'openai',
+    model: 'gpt-4',
+    inputPer1M: 30.0,
+    outputPer1M: 60.0,
+    effectiveDate: '2023-03-14',
+  },
+  {
+    provider: 'openai',
+    model: 'gpt-3.5-turbo',
+    inputPer1M: 0.5,
+    outputPer1M: 1.5,
+    effectiveDate: '2023-11-06',
+  },
+  {
+    provider: 'openai',
+    model: 'o1',
+    inputPer1M: 15.0,
+    outputPer1M: 60.0,
+    effectiveDate: '2024-12-17',
+  },
+  {
+    provider: 'openai',
+    model: 'o1-2024-12-17',
+    inputPer1M: 15.0,
+    outputPer1M: 60.0,
+    effectiveDate: '2024-12-17',
+  },
+  {
+    provider: 'openai',
+    model: 'o1-preview',
+    inputPer1M: 15.0,
+    outputPer1M: 60.0,
+    effectiveDate: '2024-09-12',
+  },
+  {
+    provider: 'openai',
+    model: 'o1-preview-2024-09-12',
+    inputPer1M: 15.0,
+    outputPer1M: 60.0,
+    effectiveDate: '2024-09-12',
+  },
+  {
+    provider: 'openai',
+    model: 'o1-mini',
+    inputPer1M: 3.0,
+    outputPer1M: 12.0,
+    effectiveDate: '2024-09-12',
+  },
+  {
+    provider: 'openai',
+    model: 'o1-mini-2024-09-12',
+    inputPer1M: 3.0,
+    outputPer1M: 12.0,
+    effectiveDate: '2024-09-12',
+  },
+
+  // Google Gemini Models
+  {
+    provider: 'gemini',
+    model: 'gemini-2.0-flash-exp',
+    inputPer1M: 0.0,
+    outputPer1M: 0.0,
+    effectiveDate: '2024-12-11',
+  },
+  {
+    provider: 'gemini',
+    model: 'gemini-2.0-flash',
+    inputPer1M: 0.075,
+    outputPer1M: 0.3,
+    effectiveDate: '2025-01-01',
+  },
+  {
+    provider: 'gemini',
+    model: 'gemini-exp-1206',
+    inputPer1M: 0.0,
+    outputPer1M: 0.0,
+    effectiveDate: '2024-12-06',
+  },
+  {
+    provider: 'gemini',
+    model: 'gemini-1.5-pro',
+    inputPer1M: 1.25,
+    outputPer1M: 5.0,
+    effectiveDate: '2024-05-14',
+  },
+  {
+    provider: 'gemini',
+    model: 'gemini-1.5-pro-002',
+    inputPer1M: 1.25,
+    outputPer1M: 5.0,
+    effectiveDate: '2024-09-24',
+  },
+  {
+    provider: 'gemini',
+    model: 'gemini-1.5-flash',
+    inputPer1M: 0.075,
+    outputPer1M: 0.3,
+    effectiveDate: '2024-05-14',
+  },
+  {
+    provider: 'gemini',
+    model: 'gemini-1.5-flash-002',
+    inputPer1M: 0.075,
+    outputPer1M: 0.3,
+    effectiveDate: '2024-09-24',
+  },
+  {
+    provider: 'gemini',
+    model: 'gemini-1.5-flash-8b',
+    inputPer1M: 0.0375,
+    outputPer1M: 0.15,
+    effectiveDate: '2024-10-03',
+  },
+
+  // Google Vertex AI (same pricing as Gemini)
+  {
+    provider: 'vertex-ai',
+    model: 'gemini-2.0-flash-exp',
+    inputPer1M: 0.0,
+    outputPer1M: 0.0,
+    effectiveDate: '2024-12-11',
+  },
+  {
+    provider: 'vertex-ai',
+    model: 'gemini-2.0-flash',
+    inputPer1M: 0.075,
+    outputPer1M: 0.3,
+    effectiveDate: '2025-01-01',
+  },
+  {
+    provider: 'vertex-ai',
+    model: 'gemini-exp-1206',
+    inputPer1M: 0.0,
+    outputPer1M: 0.0,
+    effectiveDate: '2024-12-06',
+  },
+  {
+    provider: 'vertex-ai',
+    model: 'gemini-1.5-pro',
+    inputPer1M: 1.25,
+    outputPer1M: 5.0,
+    effectiveDate: '2024-05-14',
+  },
+  {
+    provider: 'vertex-ai',
+    model: 'gemini-1.5-pro-002',
+    inputPer1M: 1.25,
+    outputPer1M: 5.0,
+    effectiveDate: '2024-09-24',
+  },
+  {
+    provider: 'vertex-ai',
+    model: 'gemini-1.5-flash',
+    inputPer1M: 0.075,
+    outputPer1M: 0.3,
+    effectiveDate: '2024-05-14',
+  },
+  {
+    provider: 'vertex-ai',
+    model: 'gemini-1.5-flash-002',
+    inputPer1M: 0.075,
+    outputPer1M: 0.3,
+    effectiveDate: '2024-09-24',
+  },
+  {
+    provider: 'vertex-ai',
+    model: 'gemini-1.5-flash-8b',
+    inputPer1M: 0.0375,
+    outputPer1M: 0.15,
+    effectiveDate: '2024-10-03',
+  },
+];
+
+/**
+ * Get pricing for a specific model
+ */
+export function getModelPricing(
+  provider: string,
+  model: string
+): ModelPricing | undefined {
+  return MODEL_PRICING.find(
+    (p) => p.provider === provider && p.model === model
+  );
+}
+
+/**
+ * Get all pricing for a provider
+ */
+export function getProviderPricing(provider: string): ModelPricing[] {
+  return MODEL_PRICING.filter((p) => p.provider === provider);
+}

package/src/pricing/types.ts

@@ -0,0 +1,32 @@
+/**
+ * Pricing and cost tracking types
+ */
+
+/**
+ * Token usage information
+ */
+export interface TokenUsage {
+  inputTokens: number;
+  outputTokens: number;
+}
+
+/**
+ * Cost estimate in USD
+ */
+export interface CostEstimate {
+  inputCost: number;
+  outputCost: number;
+  totalCost: number;
+  currency: string; // 'USD'
+}
+
+/**
+ * Model pricing per 1M tokens (USD)
+ */
+export interface ModelPricing {
+  provider: string;
+  model: string;
+  inputPer1M: number; // USD per 1M input tokens
+  outputPer1M: number; // USD per 1M output tokens
+  effectiveDate?: string; // When this pricing became effective
+}

package/src/prompts/system/base.txt

@@ -61,6 +61,48 @@ When making changes to files, first understand the file's code conventions. Mimi
 - Do not add docstrings, type annotations, or comments to code you didn't change
 - Keep solutions simple and focused. Only make changes directly requested or clearly necessary.
 
+# Asking Questions
+
+You have access to the AskUserQuestion tool to ask the user questions during execution.
+
+CRITICAL: You MUST use the AskUserQuestion tool for ALL questions with choices. NEVER write numbered lists, bullet points, or "which do you prefer" questions as plain text.
+
+## Wrong - Plain Text Questions (DO NOT DO THIS)
+
+<bad-example>
+user: Set up a new database
+assistant: I can set up a database for you. Which one would you prefer?
+1. PostgreSQL - relational database
+2. MongoDB - document database
+3. SQLite - embedded database
+</bad-example>
+
+<bad-example>
+user: Create a new project
+assistant: What type of project?
+- Web app
+- CLI tool
+- API server
+</bad-example>
+
+## Correct - Use AskUserQuestion Tool
+
+<example>
+user: Set up a new database
+assistant: [uses AskUserQuestion tool with questions: [{question: "Which database?", header: "Database", options: [{label: "PostgreSQL", description: "Relational DB"}, {label: "MongoDB", description: "Document DB"}], multiSelect: false}]]
+</example>
+
+<example>
+user: Create a new project
+assistant: [uses AskUserQuestion tool to ask about project type, framework preferences, etc.]
+</example>
+
+Use AskUserQuestion when you need to:
+- Gather user preferences (database choice, framework, etc.)
+- Clarify ambiguous instructions with specific options
+- Get decisions on implementation approaches
+- Offer choices about direction to take
+
 # Task Management
 
 You have access to the TodoWrite tool to manage tasks.

package/src/prompts/tools/ask-user.txt

@@ -0,0 +1,110 @@
+Use this tool when you need to ask the user questions during execution.
+
+CRITICAL: You MUST use this tool for ANY question with 2+ choices. NEVER present options as plain text, numbered lists, or bullet points.
+
+WRONG (never do this):
+- "Which do you prefer? 1. Option A 2. Option B"
+- "What type? - Web - CLI - API"
+- Writing any numbered or bulleted choices in your response
+
+CORRECT: Call this tool with structured questions and options.
+
+This allows you to:
+1. Gather user preferences or requirements
+2. Clarify ambiguous instructions
+3. Get decisions on implementation choices as you work
+4. Offer choices to the user about what direction to take
+
+## When to Use This Tool
+
+Use when you encounter:
+- Ambiguous requirements that could be interpreted multiple ways
+- Technology choices where user preference matters (database, framework, etc.)
+- Implementation approaches with different trade-offs
+- Features or options that should be user-selected
+
+## When NOT to Use This Tool
+
+Don't use for:
+- Asking "Is my plan ready?" or "Should I proceed?" - just proceed or use planning tools
+- Simple yes/no confirmations - just make a reasonable choice
+- Questions you can answer by reading the codebase
+- Trivial implementation details
+
+## Usage Guidelines
+
+- Keep headers short (max 12 chars): "Database", "Auth", "Features"
+- Provide 2-4 options per question
+- Put recommended option FIRST and add "(Recommended)" to its label
+- Use multiSelect: true for non-mutually-exclusive options
+- "Other" option is automatically added - don't include it manually
+- Keep option labels concise (1-5 words)
+- Descriptions should explain trade-offs or implications
+
+## Examples
+
+### Single Choice - Technology Selection
+```json
+{
+  "questions": [{
+    "question": "Which database should we use?",
+    "header": "Database",
+    "options": [
+      { "label": "PostgreSQL (Recommended)", "description": "Robust relational DB" },
+      { "label": "MongoDB", "description": "Document-based NoSQL" },
+      { "label": "SQLite", "description": "Lightweight embedded DB" }
+    ],
+    "multiSelect": false
+  }]
+}
+```
+
+### Multi-Select - Feature Selection
+```json
+{
+  "questions": [{
+    "question": "Which features should we enable?",
+    "header": "Features",
+    "options": [
+      { "label": "TypeScript", "description": "Type safety" },
+      { "label": "ESLint", "description": "Code linting" },
+      { "label": "Testing", "description": "Unit tests" },
+      { "label": "Tailwind", "description": "CSS framework" }
+    ],
+    "multiSelect": true
+  }]
+}
+```
+
+### Multiple Questions
+```json
+{
+  "questions": [
+    {
+      "question": "Which auth method?",
+      "header": "Auth",
+      "options": [
+        { "label": "OAuth 2.0 (Recommended)", "description": "Industry standard" },
+        { "label": "JWT", "description": "Stateless tokens" }
+      ],
+      "multiSelect": false
+    },
+    {
+      "question": "Which providers to support?",
+      "header": "Providers",
+      "options": [
+        { "label": "Google", "description": "Most common" },
+        { "label": "GitHub", "description": "Developer-focused" }
+      ],
+      "multiSelect": true
+    }
+  ]
+}
+```
+
+## Constraints
+
+- Maximum 4 questions per call
+- Maximum 4 options per question
+- Header: max 12 characters
+- multiSelect must be explicitly set (required, not optional)

package/src/providers/anthropic.ts

@@ -4,6 +4,7 @@
  */
 
 import Anthropic from '@anthropic-ai/sdk';
+import { calculateCost } from '../pricing/calculator.js';
 import type {
   LLMProvider,
   CompletionOptions,
@@ -45,7 +46,7 @@ export class AnthropicProvider implements LLMProvider {
       temperature: options.temperature,
     });
 
-    return this.convertResponse(response);
+    return this.convertResponse(response, options.model);
   }
 
   async *stream(options: CompletionOptions): AsyncGenerator<StreamChunk, void, unknown> {
@@ -94,15 +95,20 @@ export class AnthropicProvider implements LLMProvider {
     const finalMessage = await stream.finalMessage();
     const content = this.convertContent(finalMessage.content);
 
+    const usage = {
+      inputTokens: finalMessage.usage.input_tokens,
+      outputTokens: finalMessage.usage.output_tokens,
+    };
+
+    const cost = calculateCost(this.name, options.model, usage);
+
     yield {
       type: 'done',
       response: {
         content,
         stopReason: this.convertStopReason(finalMessage.stop_reason),
-        usage: {
-          inputTokens: finalMessage.usage.input_tokens,
-          outputTokens: finalMessage.usage.output_tokens,
-        },
+        usage,
+        cost,
       },
     };
   }
@@ -169,14 +175,19 @@ export class AnthropicProvider implements LLMProvider {
     }));
   }
 
-  private convertResponse(response: Anthropic.Message): CompletionResponse {
+  private convertResponse(response: Anthropic.Message, model: string): CompletionResponse {
+    const usage = {
+      inputTokens: response.usage.input_tokens,
+      outputTokens: response.usage.output_tokens,
+    };
+
+    const cost = calculateCost(this.name, model, usage);
+
     return {
       content: this.convertContent(response.content),
       stopReason: this.convertStopReason(response.stop_reason),
-      usage: {
-        inputTokens: response.usage.input_tokens,
-        outputTokens: response.usage.output_tokens,
-      },
+      usage,
+      cost,
     };
   }
 

package/src/providers/gemini.ts

@@ -5,6 +5,7 @@
 
 import { GoogleGenerativeAI, SchemaType } from '@google/generative-ai';
 import type { Content, Part, Tool, GenerateContentResult } from '@google/generative-ai';
+import { calculateCost } from '../pricing/calculator.js';
 import type {
   LLMProvider,
   CompletionOptions,
@@ -47,7 +48,7 @@ export class GeminiProvider implements LLMProvider {
     const contents = this.convertMessages(options.messages);
     const result = await model.generateContent({ contents });
 
-    return this.convertResponse(result);
+    return this.convertResponse(result, options.model);
   }
 
   async *stream(options: CompletionOptions): AsyncGenerator<StreamChunk, void, unknown> {
@@ -115,17 +116,22 @@ export class GeminiProvider implements LLMProvider {
     const finalResponse = await result.response;
     const stopReason = this.getStopReason(finalResponse, functionCalls.length > 0);
 
+    const usage = finalResponse.usageMetadata
+      ? {
+          inputTokens: finalResponse.usageMetadata.promptTokenCount ?? 0,
+          outputTokens: finalResponse.usageMetadata.candidatesTokenCount ?? 0,
+        }
+      : undefined;
+
+    const cost = usage ? calculateCost(this.name, options.model, usage) : undefined;
+
     yield {
       type: 'done',
       response: {
         content,
         stopReason,
-        usage: finalResponse.usageMetadata
-          ? {
-              inputTokens: finalResponse.usageMetadata.promptTokenCount ?? 0,
-              outputTokens: finalResponse.usageMetadata.candidatesTokenCount ?? 0,
-            }
-          : undefined,
+        usage,
+        cost,
       },
     };
   }
@@ -237,7 +243,7 @@ export class GeminiProvider implements LLMProvider {
     return result;
   }
 
-  private convertResponse(result: GenerateContentResult): CompletionResponse {
+  private convertResponse(result: GenerateContentResult, model: string): CompletionResponse {
     const response = result.response;
     const parts = response.candidates?.[0]?.content?.parts ?? [];
     const content: MessageContent[] = [];
@@ -261,15 +267,20 @@ export class GeminiProvider implements LLMProvider {
 
     const hasFunctionCalls = parts.some((p) => 'functionCall' in p);
 
+    const usage = response.usageMetadata
+      ? {
+          inputTokens: response.usageMetadata.promptTokenCount ?? 0,
+          outputTokens: response.usageMetadata.candidatesTokenCount ?? 0,
+        }
+      : undefined;
+
+    const cost = usage ? calculateCost(this.name, model, usage) : undefined;
+
     return {
       content,
       stopReason: this.getStopReason(response, hasFunctionCalls),
-      usage: response.usageMetadata
-        ? {
-            inputTokens: response.usageMetadata.promptTokenCount ?? 0,
-            outputTokens: response.usageMetadata.candidatesTokenCount ?? 0,
-          }
-        : undefined,
+      usage,
+      cost,
     };
   }