shark-ai 0.3.0 → 0.3.2

package/dist/bin/shark.js

@@ -5,9 +5,10 @@ import {
   colors,
   configCommand,
   loginCommand,
+  t,
   tokenStorage,
   tui
-} from "../chunk-HJYP3WB3.js";
+} from "../chunk-JQAVZJTG.js";
 
 // src/core/error/crash-handler.ts
 import fs from "fs";
@@ -694,7 +695,7 @@ async function interactiveBusinessAnalyst() {
         if (response.actions && response.actions.length > 0) {
           for (const action of response.actions) {
             if (action.type === "talk_with_user") {
-              tui.log.info(colors.green("\u{1F916} BA Agent:"));
+              tui.log.info(colors.success("\u{1F916} BA Agent:"));
               console.log(action.content);
             } else {
               tui.log.warning(`
@@ -717,9 +718,6 @@ async function interactiveBusinessAnalyst() {
             }
           }
         }
-        if (response.tokens) {
-          tui.log.info(`Tokens used: ${response.tokens.output || 0}`);
-        }
       }
     });
     tui.outro("Session complete");
@@ -1105,9 +1103,9 @@ function getAgentId3() {
 }
 async function interactiveScanAgent(options = {}) {
   FileLogger.init();
-  tui.intro("\u{1F575}\uFE0F\u200D\u2642\uFE0F  Scan Agent");
   const config = ConfigManager.getInstance().getConfig();
   const language = config.language || "English";
+  tui.intro(t("commands.scan.intro"));
   const projectRoot = process.cwd();
   let outputFile;
   if (options.output) {
@@ -1130,58 +1128,296 @@ async function interactiveScanAgent(options = {}) {
       outputFile = path4.join(outputDir, "project-context.md");
     }
   }
-  tui.log.info(`Scanning project at: ${colors.bold(projectRoot)}`);
-  tui.log.info(`Output targeted at: ${colors.bold(outputFile)}`);
-  tui.log.info(`Language: ${colors.bold(language)}`);
+  tui.log.info(`${t("commands.scan.scanningProject")} ${colors.bold(projectRoot)}`);
+  tui.log.info(`${t("commands.scan.outputTarget")} ${colors.bold(outputFile)}`);
+  tui.log.info(`${t("commands.scan.language")} ${colors.bold(language)}`);
   const configFileRelative = path4.relative(projectRoot, outputFile);
+  const initialTemplate = `# Project Context
+
+## Overview
+[TO BE ANALYZED]
+
+## Tech Stack
+[TO BE ANALYZED]
+
+## Architecture
+[TO BE ANALYZED]
+
+## Directory Structure
+[TO BE ANALYZED]
+
+## Key Components
+[TO BE ANALYZED]
+
+## API / Interfaces
+[TO BE ANALYZED]
+
+## Data Layer
+[TO BE ANALYZED]
+
+## Configuration & Environment
+[TO BE ANALYZED]
+
+## Build & Development
+[TO BE ANALYZED]
+
+## Key Patterns & Conventions
+[TO BE ANALYZED]
+`;
+  if (!fs5.existsSync(outputFile)) {
+    const BOM = "\uFEFF";
+    fs5.writeFileSync(outputFile, BOM + initialTemplate, { encoding: "utf-8" });
+    tui.log.success(`${t("commands.scan.templateCreated")} ${outputFile}`);
+  } else {
+    tui.log.info(t("commands.scan.fileExists"));
+  }
   const superPrompt = `
 You are the **Scan Agent**, an expert software architect and analyst.
-Your mission is to explore this project's codebase and generate a comprehensive context file that will be used by other AI agents (specifically a Developer Agent) to understand how to work on this project.
+Your mission is to explore this project's codebase THOROUGHLY and generate a COMPREHENSIVE context file that will be used by other AI agents (specifically a Developer Agent) to understand how to work on this project.
 
-**Goal**: Create a markdown file at: \`${configFileRelative}\`.
+**IMPORTANT**: The file \`${configFileRelative}\` has already been created with a template structure. Your job is to FILL IN each section by analyzing the project.
 
 **LANGUAGE INSTRUCTION**:
-You MUST write the content of the \`project-context.md\` file in **${language}**.
-Also, strictly interact with tools using the appropriate payload.
-
-**Instructions**:
-1.  **Analyze Structure**: Use \`list_files\` to understand the root directory and key subdirectories (src, tools, config, etc.).
-2.  **Identify Tech Stack**: Use \`read_file\` or \`search_file\` on key manifests (package.json, pom.xml, go.mod, Dockerfile, etc.) to determine languages, frameworks, and versions.
-3.  **Map Architecture**: Infer the architectural pattern (Monolith? Microservices? Clean Architecture?) based on folder structure and key files.
-4.  **Document Key Paths**: Identify where source code, tests, and configs live.
-
-**Output Format** (Markdown):
-The final action MUST be \`create_file\` (or \`modify_file\`) for the target file with the following structure:
+You MUST write the content in **${language}**.
+
+**CRITICAL STRATEGY - INCREMENTAL UPDATES**:
+DO NOT try to rewrite the entire file at once! Instead:
+
+1. **Explore** the project using \`list_files\`, \`read_file\`, \`search_file\`
+2. **Update** specific sections incrementally using \`modify_file\`
+3. **Benefit**: Build a MUCH MORE DETAILED document without context size limitations
+
+**WORKFLOW - FILL EACH SECTION:**
+
+**Step 1 - Analyze Tech Stack:**
+- \`list_files\` root directory to see project structure
+- \`read_file\` package.json (or pom.xml, go.mod, requirements.txt, etc.)
+- \`modify_file\` to replace:
+  - \`target_content\`: "## Tech Stack\\n[TO BE ANALYZED]"
+  - \`content\`: Detailed tech stack with versions, dependencies, and their purposes
+
+**Step 2 - Analyze Directory Structure:**
+- \`list_files\` on ALL key directories (src, tests, config, docs, etc.)
+- Map the complete directory tree
+- \`modify_file\` to replace:
+  - \`target_content\`: "## Directory Structure\\n[TO BE ANALYZED]"
+  - \`content\`: Visual directory tree with purpose of each folder
+
+**Step 3 - Analyze Architecture:**
+- \`read_file\` entry points (main.ts, index.js, app.py, etc.)
+- \`read_file\` 5-10 source files to understand code patterns and organization
+- Identify architectural pattern (Clean Arch, MVC, Microservices, etc.)
+- \`modify_file\` to replace:
+  - \`target_content\`: "## Architecture\\n[TO BE ANALYZED]"
+  - \`content\`: Comprehensive architectural description with patterns and module organization
+
+**Step 4 - Document Components:**
+- Identify ALL major modules/components by reading source files
+- For each component: location, purpose, key files
+- \`modify_file\` to replace:
+  - \`target_content\`: "## Key Components\\n[TO BE ANALYZED]"
+  - \`content\`: Detailed list of components with their purposes
+
+**Step 5 - Document APIs (if applicable):**
+- Search for route definitions, controllers, API endpoints
+- Document base URL, main endpoints, request/response patterns
+- \`modify_file\` to replace:
+  - \`target_content\`: "## API / Interfaces\\n[TO BE ANALYZED]"
+  - \`content\`: API documentation (or "Not applicable" if no API)
+
+**Step 6 - Document Data Layer (if applicable):**
+- Search for database configs, ORM setup, model definitions
+- Document database type, ORM tool, key entities/tables
+- \`modify_file\` to replace:
+  - \`target_content\`: "## Data Layer\\n[TO BE ANALYZED]"
+  - \`content\`: Data layer details (or "Not applicable" if no database)
+
+**Step 7 - Configuration & Environment:**
+- Read config files (.env.example, config/, etc.)
+- Identify environment variables and their purposes
+- \`modify_file\` to replace:
+  - \`target_content\`: "## Configuration & Environment\\n[TO BE ANALYZED]"
+  - \`content\`: List of env vars and config files
+
+**Step 8 - Build & Development:**
+- Read package.json scripts, Makefile, build configs
+- Document dev, build, test, lint commands
+- \`modify_file\` to replace:
+  - \`target_content\`: "## Build & Development\\n[TO BE ANALYZED]"
+  - \`content\`: Commands for development workflow
+
+**Step 9 - Patterns & Conventions:**
+- Based on files read, document naming conventions, code organization
+- Note error handling, logging strategies
+- \`modify_file\` to replace:
+  - \`target_content\`: "## Key Patterns & Conventions\\n[TO BE ANALYZED]"
+  - \`content\`: Observed patterns and conventions
+
+**Step 10 - Overview (LAST):**
+- Synthesize all findings into a comprehensive overview
+- \`modify_file\` to replace:
+  - \`target_content\`: "## Overview\\n[TO BE ANALYZED]"
+  - \`content\`: Detailed project description with purpose and main functionality
+
+**HOW TO USE modify_file:**
+\`\`\`json
+{
+  "actions": [{
+    "type": "modify_file",
+    "path": "${configFileRelative}",
+    "target_content": "## Tech Stack\\n[TO BE ANALYZED]",
+    "content": "## Tech Stack\\n- **Language**: TypeScript 5.3\\n- **Runtime**: Node.js 20.x\\n..."
+  }]
+}
+\`\`\`
 
+\`\`\`markdown
 # Project Context
 
 ## Overview
-[Brief description of what this project seems to be]
+[TO BE ANALYZED]
 
 ## Tech Stack
-- **Language**: [e.g. TypeScript]
-- **Framework**: [e.g. React, Express, NestJS]
-- **Build Tool**: [e.g. Vite, Webpack]
-- **Database**: [e.g. PostgreSQL, Prisma] (if detected)
+[TO BE ANALYZED]
 
 ## Architecture
-[Description of the folder structure and architectural patterns detected]
+[TO BE ANALYZED]
 
-## Key Locations
-- **Source**: [path/to/src]
-- **Tests**: [path/to/tests]
-- **Config**: [path/to/config]
+## Directory Structure
+[TO BE ANALYZED]
 
-## Commands
-[List of discovered npm scripts or makefile commands for dev, build, test]
+## Key Components
+[TO BE ANALYZED]
 
----
+## API / Interfaces
+[TO BE ANALYZED]
+
+## Data Layer
+[TO BE ANALYZED]
+
+## Configuration & Environment
+[TO BE ANALYZED]
+
+## Build & Development
+[TO BE ANALYZED]
+
+## Key Patterns & Conventions
+[TO BE ANALYZED]
+\`\`\`
+
+**Step 2 - Explore and Update Incrementally:**
+After creating the template, perform these analyses and UPDATE each section:
+
+**2.1 - Analyze Tech Stack:**
+- \`list_files\` root directory
+- \`read_file\` package.json (or equivalent manifest)
+- \`modify_file\` to replace "## Tech Stack\\n[TO BE ANALYZED]" with detailed findings
+
+**2.2 - Analyze Directory Structure:**
+- \`list_files\` on key directories (src, tests, config, etc.)
+- \`modify_file\` to replace "## Directory Structure\\n[TO BE ANALYZED]" with complete structure
+
+**2.3 - Analyze Architecture:**
+- \`read_file\` entry points (main.ts, index.js, etc.)
+- \`read_file\` 5-10 source files to understand patterns
+- \`modify_file\` to replace "## Architecture\\n[TO BE ANALYZED]" with architectural insights
 
-**Rules**:
-- Do NOT guess. If you are unsure, check the file.
-- Be concise.
-- Focus on FACTS that a Developer Agent needs to know to write code correcty.
-- Start by listing the root directory.
+**2.4 - Document Components:**
+- Identify major modules/components
+- \`modify_file\` to replace "## Key Components\\n[TO BE ANALYZED]" with component details
+
+**2.5 - Document APIs (if applicable):**
+- Search for route definitions, controllers, API endpoints
+- \`modify_file\` to replace "## API / Interfaces\\n[TO BE ANALYZED]"
+
+**2.6 - Document Data Layer (if applicable):**
+- Search for database configs, ORM, models
+- \`modify_file\` to replace "## Data Layer\\n[TO BE ANALYZED]"
+
+**2.7 - Final Touches:**
+- Update remaining sections (Config, Build, Patterns)
+- Ensure Overview is comprehensive
+
+**HOW TO USE modify_file:**
+\`\`\`json
+{
+  "actions": [{
+    "type": "modify_file",
+    "path": "${configFileRelative}",
+    "target_content": "## Tech Stack\\n[TO BE ANALYZED]",
+    "content": "## Tech Stack\\n- **Language**: TypeScript 5.3\\n- **Runtime**: Node.js 20.x\\n- **Framework**: Express 4.18\\n..."
+  }]
+}
+\`\`\`
+
+**SECTION TEMPLATES (For your updates):**
+
+**Tech Stack:**
+\`\`\`markdown
+## Tech Stack
+- **Language**: [name + version]
+- **Runtime**: [name + version]
+- **Framework**: [name + version]
+- **Build Tool**: [name]
+- **Testing**: [framework]
+- **Key Dependencies**:
+  - [dep-name]: [purpose]
+  - [dep-name]: [purpose]
+\`\`\`
+
+**Architecture:**
+\`\`\`markdown
+## Architecture
+[Comprehensive description]
+- **Pattern**: [Clean Arch, MVC, etc.]
+- **Module Organization**: [how modules are structured]
+- **Layer Separation**: [controllers, services, repos]
+- **Configuration**: [how config is managed]
+\`\`\`
+
+**Directory Structure:**
+\`\`\`markdown
+## Directory Structure
+\\\`\\\`\\\`
+/src
+  /core        - [Purpose]
+  /commands    - [Purpose]
+  /ui          - [Purpose]
+/tests         - [Purpose]
+/docs          - [Purpose]
+\\\`\\\`\\\`
+\`\`\`
+
+**Key Components:**
+\`\`\`markdown
+## Key Components
+
+### [Component Name 1]
+- **Location**: \\\`path/to/component\\\`
+- **Purpose**: [What it does]
+- **Key Files**: [Important files]
+
+### [Component Name 2]
+- **Location**: \\\`path/to/component\\\`
+- **Purpose**: [What it does]
+- **Key Files**: [Important files]
+\`\`\`
+
+**DEPTH REQUIREMENT**:
+- Read MULTIPLE files (5-10 minimum) to understand patterns
+- Identify ALL major modules/components
+- Document API routes, database schemas if applicable
+- Note design patterns and architectural decisions
+- List important dependencies with their purposes
+
+**RULES**:
+- Create template FIRST (step 1)
+- Update sections INCREMENTALLY (steps 2-7)
+- Do NOT wait to write everything at the end
+- Use \`modify_file\` to replace "[TO BE ANALYZED]" sections
+- Be DETAILED and COMPREHENSIVE
+- Take your time - you have up to 30 steps
+- Verify facts with \`read_file\` or \`search_file\` before documenting
 `.trim();
   await runScanLoop(superPrompt, outputFile);
 }
@@ -1189,38 +1425,39 @@ async function runScanLoop(initialPrompt, targetPath) {
   let nextPrompt = initialPrompt;
   let keepGoing = true;
   let stepCount = 0;
-  const MAX_STEPS = 15;
+  const MAX_STEPS = 60;
   while (keepGoing && stepCount < MAX_STEPS) {
     stepCount++;
     const spinner = tui.spinner();
-    spinner.start(`\u{1F575}\uFE0F\u200D\u2642\uFE0F  Scan Agent analyzing (Step ${stepCount}/${MAX_STEPS})...`);
+    const msg = t("commands.scan.analyzing").replace("{step}", stepCount.toString());
+    spinner.start(msg);
     let responseText = "";
     let lastResponse = null;
     try {
       lastResponse = await callScanAgentApi(nextPrompt, (chunk) => {
         responseText += chunk;
       });
-      spinner.stop("Step complete");
-      if (lastResponse && lastResponse.actions) {
+      spinner.stop(t("commands.scan.stepComplete"));
+      if (lastResponse && lastResponse.actions && lastResponse.actions.length > 0) {
         let executionResults = "";
         let fileCreated = false;
         for (const action of lastResponse.actions) {
           if (action.type === "list_files") {
-            tui.log.info(`\u{1F4C2} Scanning dir: ${colors.bold(action.path || ".")}`);
+            tui.log.info(t("commands.scan.scanningDir").replace("{0}", colors.bold(action.path || ".")));
             const result = handleListFiles(action.path || ".");
             executionResults += `[Action list_files(${action.path}) Result]:
 ${result}
 
 `;
           } else if (action.type === "read_file") {
-            tui.log.info(`\u{1F4D6} Reading file: ${colors.bold(action.path || "")}`);
+            tui.log.info(t("commands.scan.readingFile").replace("{0}", colors.bold(action.path || "")));
             const result = handleReadFile(action.path || "");
             executionResults += `[Action read_file(${action.path}) Result]:
 ${result}
 
 `;
           } else if (action.type === "search_file") {
-            tui.log.info(`\u{1F50D} Searching: ${colors.bold(action.path || "")}`);
+            tui.log.info(t("commands.scan.searching").replace("{0}", colors.bold(action.path || "")));
             const result = handleSearchFile(action.path || "");
             executionResults += `[Action search_file(${action.path}) Result]:
 ${result}
@@ -1231,52 +1468,87 @@ ${result}
             const resolvedTargetPath = path4.resolve(targetPath);
             let isTarget = resolvedActionPath === resolvedTargetPath;
             if (!isTarget && path4.basename(action.path || "") === "project-context.md") {
-              tui.log.warning(`Agent targeted '${action.path}' but we enforce '${path4.relative(process.cwd(), targetPath)}'. Redirecting write.`);
+              tui.log.warning(t("commands.scan.targetRedirect").replace("{0}", action.path || "").replace("{1}", path4.relative(process.cwd(), targetPath)));
               isTarget = true;
               action.path = targetPath;
             }
             if (isTarget) {
               const finalPath = targetPath;
+              const BOM = "\uFEFF";
               if (action.type === "create_file") {
-                fs5.writeFileSync(finalPath, action.content || "");
-                tui.log.success(`\u2705 Generated Context: ${finalPath}`);
+                const contentToWrite = action.content || "";
+                const finalContent = contentToWrite.startsWith(BOM) ? contentToWrite : BOM + contentToWrite;
+                fs5.writeFileSync(finalPath, finalContent, { encoding: "utf-8" });
+                tui.log.success(t("commands.scan.generated").replace("{0}", finalPath));
                 fileCreated = true;
               } else {
-                fs5.writeFileSync(finalPath, action.content || "");
-                tui.log.success(`\u2705 Updated Context: ${finalPath}`);
-                fileCreated = true;
+                if (fs5.existsSync(finalPath)) {
+                  const currentContent = fs5.readFileSync(finalPath, "utf-8");
+                  if (action.target_content && currentContent.includes(action.target_content)) {
+                    const newContent = currentContent.replace(action.target_content, action.content || "");
+                    const finalContent = newContent.startsWith(BOM) ? newContent : BOM + newContent;
+                    fs5.writeFileSync(finalPath, finalContent, { encoding: "utf-8" });
+                    tui.log.success(t("commands.scan.updated").replace("{0}", finalPath));
+                    fileCreated = true;
+                  } else {
+                    tui.log.warning(t("commands.scan.error") + ": " + t("commands.scan.contentNotFound"));
+                    executionResults += `[Action ${action.type}]: Failed. Target content not found in file.
+`;
+                    fileCreated = false;
+                  }
+                } else {
+                  tui.log.warning(t("commands.scan.error") + ": " + t("commands.scan.notFound"));
+                }
               }
               executionResults += `[Action ${action.type}]: Success. Task Completed.
 `;
             } else {
-              tui.log.warning(`Agent wants to write to unexpected file: ${action.path}`);
-              executionResults += `[Action ${action.type}]: Skipped (Scan Agent only writes context file)
+              tui.log.warning(t("commands.scan.error"));
+              executionResults += `[Action ${action.type}]: ${t("commands.scan.skipped")}
 `;
             }
           } else if (action.type === "talk_with_user") {
-            tui.log.info(colors.primary("\u{1F916} Scan Agent asks:"));
+            tui.log.info(colors.primary(t("commands.scan.agentAsks")));
             console.log(action.content);
-            const reply = await tui.text({ message: "Agent needs input:", placeholder: "Reply..." });
+            const reply = await tui.text({ message: t("commands.scan.agentInput"), placeholder: t("commands.scan.replyPlaceholder") });
             executionResults += `[User Reply]: ${reply}
 `;
           }
         }
         if (fileCreated) {
-          tui.log.success("\u2728 Scan completed successfully!");
-          keepGoing = false;
+          const currentContent = fs5.readFileSync(targetPath, "utf-8");
+          const pendingSections = [];
+          const lines = currentContent.split("\n");
+          let currentSection = "";
+          for (const line of lines) {
+            if (line.startsWith("## ")) {
+              currentSection = line.substring(3).trim();
+            }
+            if (line.includes("[TO BE ANALYZED]")) {
+              if (currentSection && !pendingSections.includes(currentSection)) {
+                pendingSections.push(currentSection);
+              }
+            }
+          }
+          const pendingMsg = pendingSections.length > 0 ? `
+
+[System Helper]: ${t("commands.scan.pendingSections").replace("{0}", pendingSections.join(", "))}` : `
+
+[System Helper]: ${t("commands.scan.allPopulated")}`;
+          nextPrompt = `${executionResults}
+
+[System]: File updated successfully.${pendingMsg} Please continue with the next step of the analysis and focus on the pending sections.`;
+          FileLogger.log("SCAN", "Section updated, continuing loop", { step: stepCount, pending: pendingSections.length });
         } else {
           nextPrompt = executionResults;
           FileLogger.log("SCAN", "Auto-replying with results", { length: executionResults.length });
         }
       } else {
-        if (stepCount > 1) {
-          tui.log.warning("Scan Agent stopped without actions.");
-          keepGoing = false;
-        } else {
-        }
+        tui.log.success(t("commands.scan.completed"));
+        keepGoing = false;
       }
     } catch (error) {
-      spinner.stop("Error");
+      spinner.stop(t("common.error"));
       tui.log.error(error.message);
       keepGoing = false;
     }