gitnexus 1.2.3 → 1.2.5

package/dist/cli/index.js

@@ -53,9 +53,10 @@ program
     .command('wiki [path]')
     .description('Generate repository wiki from knowledge graph')
     .option('-f, --force', 'Force full regeneration even if up to date')
-    .option('--model <model>', 'LLM model name (default: gpt-4o-mini)')
+    .option('--model <model>', 'LLM model name (default: minimax/minimax-m2.5)')
     .option('--base-url <url>', 'LLM API base URL (default: OpenAI)')
     .option('--api-key <key>', 'LLM API key (saved to ~/.gitnexus/config.json)')
+    .option('--concurrency <n>', 'Parallel LLM calls (default: 3)', '3')
     .option('--gist', 'Publish wiki as a public GitHub Gist after generation')
     .action(wikiCommand);
 program

package/dist/cli/wiki.d.ts

@@ -9,6 +9,7 @@ export interface WikiCommandOptions {
     model?: string;
     baseUrl?: string;
     apiKey?: string;
+    concurrency?: string;
     gist?: boolean;
 }
 export declare const wikiCommand: (inputPath?: string, options?: WikiCommandOptions) => Promise<void>;

package/dist/cli/wiki.js

@@ -141,7 +141,7 @@ export const wikiCommand = async (inputPath, options) => {
             let defaultModel;
             if (choice === '2') {
                 baseUrl = 'https://openrouter.ai/api/v1';
-                defaultModel = 'openai/gpt-4o-mini';
+                defaultModel = 'minimax/minimax-m2.5';
             }
             else if (choice === '3') {
                 baseUrl = await prompt('  Base URL (e.g. http://localhost:11434/v1): ');
@@ -186,7 +186,7 @@ export const wikiCommand = async (inputPath, options) => {
             llmConfig = { ...llmConfig, apiKey: key, baseUrl, model };
         }
     }
-    // ── Setup progress bar ──────────────────────────────────────────────
+    // ── Setup progress bar with elapsed timer ──────────────────────────
     const bar = new cliProgress.SingleBar({
         format: '  {bar} {percentage}% | {phase}',
         barCompleteChar: '\u2588',
@@ -199,17 +199,35 @@ export const wikiCommand = async (inputPath, options) => {
     }, cliProgress.Presets.shades_grey);
     bar.start(100, 0, { phase: 'Initializing...' });
     const t0 = Date.now();
+    let lastPhase = '';
+    let phaseStart = t0;
+    // Tick elapsed time every second while stuck on the same phase
+    const elapsedTimer = setInterval(() => {
+        if (lastPhase) {
+            const elapsed = Math.round((Date.now() - phaseStart) / 1000);
+            if (elapsed >= 3) {
+                bar.update({ phase: `${lastPhase} (${elapsed}s)` });
+            }
+        }
+    }, 1000);
     // ── Run generator ───────────────────────────────────────────────────
     const wikiOptions = {
         force: options?.force,
         model: options?.model,
         baseUrl: options?.baseUrl,
+        concurrency: options?.concurrency ? parseInt(options.concurrency, 10) : undefined,
     };
     const generator = new WikiGenerator(repoPath, storagePath, kuzuPath, llmConfig, wikiOptions, (phase, percent, detail) => {
-        bar.update(percent, { phase: detail || phase });
+        const label = detail || phase;
+        if (label !== lastPhase) {
+            lastPhase = label;
+            phaseStart = Date.now();
+        }
+        bar.update(percent, { phase: label });
     });
     try {
         const result = await generator.run();
+        clearInterval(elapsedTimer);
         bar.update(100, { phase: 'Done' });
         bar.stop();
         const elapsed = ((Date.now() - t0) / 1000).toFixed(1);
@@ -237,6 +255,7 @@ export const wikiCommand = async (inputPath, options) => {
         await maybePublishGist(viewerPath, options?.gist);
     }
     catch (err) {
+        clearInterval(elapsedTimer);
         bar.stop();
         if (err.message?.includes('No source files')) {
             console.log(`\n  ${err.message}\n`);

package/dist/core/wiki/generator.d.ts

@@ -16,6 +16,7 @@ export interface WikiOptions {
     baseUrl?: string;
     apiKey?: string;
     maxTokensPerModule?: number;
+    concurrency?: number;
 }
 export interface WikiMeta {
     fromCommit: string;
@@ -38,10 +39,17 @@ export declare class WikiGenerator {
     private kuzuPath;
     private llmConfig;
     private maxTokensPerModule;
+    private concurrency;
     private options;
     private onProgress;
     private failedModules;
     constructor(repoPath: string, storagePath: string, kuzuPath: string, llmConfig: LLMConfig, options?: WikiOptions, onProgress?: ProgressCallback);
+    private lastPercent;
+    /**
+     * Create streaming options that report LLM progress to the progress bar.
+     * Uses the last known percent so streaming doesn't reset the bar backwards.
+     */
+    private streamOpts;
     /**
      * Main entry point. Runs the full pipeline or incremental update.
      */
@@ -65,11 +73,6 @@ export declare class WikiGenerator {
      * Split a large module into sub-modules by subdirectory.
      */
     private splitBySubdirectory;
-    /**
-     * Recursively generate pages for a module tree node.
-     * Returns count of pages generated.
-     */
-    private generateModulePage;
     /**
      * Generate a leaf module page from source code + graph data.
      */
@@ -88,6 +91,16 @@ export declare class WikiGenerator {
     private readProjectInfo;
     private extractModuleFiles;
     private countModules;
+    /**
+     * Flatten the module tree into leaf nodes and parent nodes.
+     * Leaves can be processed in parallel; parents must wait for children.
+     */
+    private flattenModuleTree;
+    /**
+     * Run async tasks in parallel with a concurrency limit and adaptive rate limiting.
+     * If a 429 rate limit is hit, concurrency is temporarily reduced.
+     */
+    private runParallel;
     private findNodeBySlug;
     private slugify;
     private fileExists;

package/dist/core/wiki/generator.js

@@ -28,6 +28,7 @@ export class WikiGenerator {
     kuzuPath;
     llmConfig;
     maxTokensPerModule;
+    concurrency;
     options;
     onProgress;
     failedModules = [];
@@ -39,7 +40,27 @@ export class WikiGenerator {
         this.options = options;
         this.llmConfig = llmConfig;
         this.maxTokensPerModule = options.maxTokensPerModule ?? DEFAULT_MAX_TOKENS_PER_MODULE;
-        this.onProgress = onProgress || (() => { });
+        this.concurrency = options.concurrency ?? 3;
+        const progressFn = onProgress || (() => { });
+        this.onProgress = (phase, percent, detail) => {
+            if (percent > 0)
+                this.lastPercent = percent;
+            progressFn(phase, percent, detail);
+        };
+    }
+    lastPercent = 0;
+    /**
+     * Create streaming options that report LLM progress to the progress bar.
+     * Uses the last known percent so streaming doesn't reset the bar backwards.
+     */
+    streamOpts(label, fixedPercent) {
+        return {
+            onChunk: (chars) => {
+                const tokens = Math.round(chars / 4);
+                const pct = fixedPercent ?? this.lastPercent;
+                this.onProgress('stream', pct, `${label} (${tokens} tok)`);
+            },
+        };
     }
     /**
      * Main entry point. Runs the full pipeline or incremental update.
@@ -123,16 +144,54 @@ export class WikiGenerator {
         // Phase 1: Build module tree
         const moduleTree = await this.buildModuleTree(enrichedFiles);
         pagesGenerated = 0;
-        // Phase 2: Generate module pages (bottom-up)
+        // Phase 2: Generate module pages (parallel with concurrency limit)
         const totalModules = this.countModules(moduleTree);
         let modulesProcessed = 0;
-        for (const node of moduleTree) {
-            const generated = await this.generateModulePage(node, () => {
-                modulesProcessed++;
-                const percent = 30 + Math.round((modulesProcessed / totalModules) * 55);
-                this.onProgress('modules', percent, `${modulesProcessed}/${totalModules} modules`);
-            });
-            pagesGenerated += generated;
+        const reportProgress = (moduleName) => {
+            modulesProcessed++;
+            const percent = 30 + Math.round((modulesProcessed / totalModules) * 55);
+            const detail = moduleName
+                ? `${modulesProcessed}/${totalModules} — ${moduleName}`
+                : `${modulesProcessed}/${totalModules} modules`;
+            this.onProgress('modules', percent, detail);
+        };
+        // Flatten tree into layers: leaves first, then parents
+        // Leaves can run in parallel; parents must wait for their children
+        const { leaves, parents } = this.flattenModuleTree(moduleTree);
+        // Process all leaf modules in parallel
+        pagesGenerated += await this.runParallel(leaves, async (node) => {
+            const pagePath = path.join(this.wikiDir, `${node.slug}.md`);
+            if (await this.fileExists(pagePath)) {
+                reportProgress(node.name);
+                return 0;
+            }
+            try {
+                await this.generateLeafPage(node);
+                reportProgress(node.name);
+                return 1;
+            }
+            catch (err) {
+                this.failedModules.push(node.name);
+                reportProgress(`Failed: ${node.name}`);
+                return 0;
+            }
+        });
+        // Process parent modules sequentially (they depend on child docs)
+        for (const node of parents) {
+            const pagePath = path.join(this.wikiDir, `${node.slug}.md`);
+            if (await this.fileExists(pagePath)) {
+                reportProgress(node.name);
+                continue;
+            }
+            try {
+                await this.generateParentPage(node);
+                pagesGenerated++;
+                reportProgress(node.name);
+            }
+            catch (err) {
+                this.failedModules.push(node.name);
+                reportProgress(`Failed: ${node.name}`);
+            }
         }
         // Phase 3: Generate overview
         this.onProgress('overview', 88, 'Generating overview page...');
@@ -174,7 +233,7 @@ export class WikiGenerator {
             FILE_LIST: fileList,
             DIRECTORY_TREE: dirTree,
         });
-        const response = await callLLM(prompt, this.llmConfig, GROUPING_SYSTEM_PROMPT);
+        const response = await callLLM(prompt, this.llmConfig, GROUPING_SYSTEM_PROMPT, this.streamOpts('Grouping files', 15));
         const grouping = this.parseGroupingResponse(response.content, files);
         // Convert to tree nodes
         const tree = [];
@@ -284,53 +343,6 @@ export class WikiGenerator {
         }));
     }
     // ─── Phase 2: Generate Module Pages ─────────────────────────────────
-    /**
-     * Recursively generate pages for a module tree node.
-     * Returns count of pages generated.
-     */
-    async generateModulePage(node, onPageDone) {
-        let count = 0;
-        // If node has children, generate children first (bottom-up)
-        if (node.children && node.children.length > 0) {
-            for (const child of node.children) {
-                count += await this.generateModulePage(child, onPageDone);
-            }
-            // Then generate parent page from children docs
-            const pagePath = path.join(this.wikiDir, `${node.slug}.md`);
-            // Resumability: skip if page already exists
-            if (await this.fileExists(pagePath)) {
-                onPageDone();
-                return count;
-            }
-            try {
-                await this.generateParentPage(node);
-                count++;
-            }
-            catch (err) {
-                this.failedModules.push(node.name);
-                this.onProgress('modules', 0, `Failed: ${node.name} — ${err.message?.slice(0, 80)}`);
-            }
-            onPageDone();
-            return count;
-        }
-        // Leaf module — generate from source code
-        const pagePath = path.join(this.wikiDir, `${node.slug}.md`);
-        // Resumability: skip if page already exists
-        if (await this.fileExists(pagePath)) {
-            onPageDone();
-            return count;
-        }
-        try {
-            await this.generateLeafPage(node);
-            count++;
-        }
-        catch (err) {
-            this.failedModules.push(node.name);
-            this.onProgress('modules', 0, `Failed: ${node.name} — ${err.message?.slice(0, 80)}`);
-        }
-        onPageDone();
-        return count;
-    }
     /**
      * Generate a leaf module page from source code + graph data.
      */
@@ -358,7 +370,7 @@ export class WikiGenerator {
             INCOMING_CALLS: formatCallEdges(interCalls.incoming),
             PROCESSES: formatProcesses(processes),
         });
-        const response = await callLLM(prompt, this.llmConfig, MODULE_SYSTEM_PROMPT);
+        const response = await callLLM(prompt, this.llmConfig, MODULE_SYSTEM_PROMPT, this.streamOpts(node.name));
         // Write page with front matter
         const pageContent = `# ${node.name}\n\n${response.content}`;
         await fs.writeFile(path.join(this.wikiDir, `${node.slug}.md`), pageContent, 'utf-8');
@@ -394,7 +406,7 @@ export class WikiGenerator {
             CROSS_MODULE_CALLS: formatCallEdges(crossCalls),
             CROSS_PROCESSES: formatProcesses(processes),
         });
-        const response = await callLLM(prompt, this.llmConfig, PARENT_SYSTEM_PROMPT);
+        const response = await callLLM(prompt, this.llmConfig, PARENT_SYSTEM_PROMPT, this.streamOpts(node.name));
         const pageContent = `# ${node.name}\n\n${response.content}`;
         await fs.writeFile(path.join(this.wikiDir, `${node.slug}.md`), pageContent, 'utf-8');
     }
@@ -430,7 +442,7 @@ export class WikiGenerator {
             MODULE_EDGES: edgesText,
             TOP_PROCESSES: formatProcesses(topProcesses),
         });
-        const response = await callLLM(prompt, this.llmConfig, OVERVIEW_SYSTEM_PROMPT);
+        const response = await callLLM(prompt, this.llmConfig, OVERVIEW_SYSTEM_PROMPT, this.streamOpts('Generating overview', 88));
         const pageContent = `# ${path.basename(this.repoPath)} — Wiki\n\n${response.content}`;
         await fs.writeFile(path.join(this.wikiDir, 'overview.md'), pageContent, 'utf-8');
     }
@@ -484,26 +496,43 @@ export class WikiGenerator {
             existingMeta.moduleFiles['Other'].push(...newFiles);
             affectedModules.add('Other');
         }
-        // Regenerate affected module pages
+        // Regenerate affected module pages (parallel)
         let pagesGenerated = 0;
         const moduleTree = existingMeta.moduleTree;
         const affectedArray = Array.from(affectedModules);
         this.onProgress('incremental', 20, `Regenerating ${affectedArray.length} module(s)...`);
-        for (let i = 0; i < affectedArray.length; i++) {
-            const modSlug = this.slugify(affectedArray[i]);
+        const affectedNodes = [];
+        for (const mod of affectedArray) {
+            const modSlug = this.slugify(mod);
             const node = this.findNodeBySlug(moduleTree, modSlug);
             if (node) {
-                // Delete existing page to force re-generation
                 try {
                     await fs.unlink(path.join(this.wikiDir, `${node.slug}.md`));
                 }
                 catch { }
-                await this.generateModulePage(node, () => { });
-                pagesGenerated++;
+                affectedNodes.push(node);
             }
-            const percent = 20 + Math.round(((i + 1) / affectedArray.length) * 60);
-            this.onProgress('incremental', percent, `${i + 1}/${affectedArray.length} modules`);
         }
+        let incProcessed = 0;
+        pagesGenerated += await this.runParallel(affectedNodes, async (node) => {
+            try {
+                if (node.children && node.children.length > 0) {
+                    await this.generateParentPage(node);
+                }
+                else {
+                    await this.generateLeafPage(node);
+                }
+                incProcessed++;
+                const percent = 20 + Math.round((incProcessed / affectedNodes.length) * 60);
+                this.onProgress('incremental', percent, `${incProcessed}/${affectedNodes.length} — ${node.name}`);
+                return 1;
+            }
+            catch (err) {
+                this.failedModules.push(node.name);
+                incProcessed++;
+                return 0;
+            }
+        });
         // Regenerate overview if any pages changed
         if (pagesGenerated > 0) {
             this.onProgress('incremental', 85, 'Updating overview...');
@@ -637,6 +666,80 @@ export class WikiGenerator {
         }
         return count;
     }
+    /**
+     * Flatten the module tree into leaf nodes and parent nodes.
+     * Leaves can be processed in parallel; parents must wait for children.
+     */
+    flattenModuleTree(tree) {
+        const leaves = [];
+        const parents = [];
+        for (const node of tree) {
+            if (node.children && node.children.length > 0) {
+                for (const child of node.children) {
+                    leaves.push(child);
+                }
+                parents.push(node);
+            }
+            else {
+                leaves.push(node);
+            }
+        }
+        return { leaves, parents };
+    }
+    /**
+     * Run async tasks in parallel with a concurrency limit and adaptive rate limiting.
+     * If a 429 rate limit is hit, concurrency is temporarily reduced.
+     */
+    async runParallel(items, fn) {
+        let total = 0;
+        let activeConcurrency = this.concurrency;
+        let running = 0;
+        let idx = 0;
+        return new Promise((resolve, reject) => {
+            const next = () => {
+                while (running < activeConcurrency && idx < items.length) {
+                    const item = items[idx++];
+                    running++;
+                    fn(item)
+                        .then((count) => {
+                        total += count;
+                        running--;
+                        if (idx >= items.length && running === 0) {
+                            resolve(total);
+                        }
+                        else {
+                            next();
+                        }
+                    })
+                        .catch((err) => {
+                        running--;
+                        // On rate limit, reduce concurrency temporarily
+                        if (err.message?.includes('429')) {
+                            activeConcurrency = Math.max(1, activeConcurrency - 1);
+                            this.onProgress('modules', this.lastPercent, `Rate limited — concurrency → ${activeConcurrency}`);
+                            // Re-queue the item
+                            idx--;
+                            setTimeout(next, 5000);
+                        }
+                        else {
+                            if (idx >= items.length && running === 0) {
+                                resolve(total);
+                            }
+                            else {
+                                next();
+                            }
+                        }
+                    });
+                }
+            };
+            if (items.length === 0) {
+                resolve(0);
+            }
+            else {
+                next();
+            }
+        });
+    }
     findNodeBySlug(tree, slug) {
         for (const node of tree) {
             if (node.slug === slug)

package/dist/core/wiki/llm-client.d.ts

@@ -29,8 +29,12 @@ export declare function resolveLLMConfig(overrides?: Partial<LLMConfig>): Promis
  * Estimate token count from text (rough heuristic: ~4 chars per token).
  */
 export declare function estimateTokens(text: string): number;
+export interface CallLLMOptions {
+    onChunk?: (charsReceived: number) => void;
+}
 /**
  * Call an OpenAI-compatible LLM API.
- * Retries once on transient failures (5xx, network errors).
+ * Uses streaming when onChunk callback is provided for real-time progress.
+ * Retries up to 3 times on transient failures (429, 5xx, network errors).
  */
-export declare function callLLM(prompt: string, config: LLMConfig, systemPrompt?: string): Promise<LLMResponse>;
+export declare function callLLM(prompt: string, config: LLMConfig, systemPrompt?: string, options?: CallLLMOptions): Promise<LLMResponse>;

package/dist/core/wiki/llm-client.js

@@ -25,11 +25,11 @@ export async function resolveLLMConfig(overrides) {
         baseUrl: overrides?.baseUrl
             || process.env.GITNEXUS_LLM_BASE_URL
             || savedConfig.baseUrl
-            || 'https://api.openai.com/v1',
+            || 'https://openrouter.ai/api/v1',
         model: overrides?.model
             || process.env.GITNEXUS_MODEL
             || savedConfig.model
-            || 'gpt-4o-mini',
+            || 'minimax/minimax-m2.5',
         maxTokens: overrides?.maxTokens ?? 16_384,
         temperature: overrides?.temperature ?? 0,
     };
@@ -42,23 +42,28 @@ export function estimateTokens(text) {
 }
 /**
  * Call an OpenAI-compatible LLM API.
- * Retries once on transient failures (5xx, network errors).
+ * Uses streaming when onChunk callback is provided for real-time progress.
+ * Retries up to 3 times on transient failures (429, 5xx, network errors).
  */
-export async function callLLM(prompt, config, systemPrompt) {
+export async function callLLM(prompt, config, systemPrompt, options) {
     const messages = [];
     if (systemPrompt) {
         messages.push({ role: 'system', content: systemPrompt });
     }
     messages.push({ role: 'user', content: prompt });
     const url = `${config.baseUrl.replace(/\/+$/, '')}/chat/completions`;
+    const useStream = !!options?.onChunk;
     const body = {
         model: config.model,
         messages,
         max_tokens: config.maxTokens,
         temperature: config.temperature,
     };
+    if (useStream)
+        body.stream = true;
+    const MAX_RETRIES = 3;
     let lastError = null;
-    for (let attempt = 0; attempt < 2; attempt++) {
+    for (let attempt = 0; attempt < MAX_RETRIES; attempt++) {
         try {
             const response = await fetch(url, {
                 method: 'POST',
@@ -70,19 +75,25 @@ export async function callLLM(prompt, config, systemPrompt) {
             });
             if (!response.ok) {
                 const errorText = await response.text().catch(() => 'unknown error');
-                // Rate limit — wait and retry
-                if (response.status === 429 && attempt === 0) {
-                    const retryAfter = parseInt(response.headers.get('retry-after') || '5', 10);
-                    await sleep(retryAfter * 1000);
+                // Rate limit — wait with exponential backoff and retry
+                if (response.status === 429 && attempt < MAX_RETRIES - 1) {
+                    const retryAfter = parseInt(response.headers.get('retry-after') || '0', 10);
+                    const delay = retryAfter > 0 ? retryAfter * 1000 : (2 ** attempt) * 3000;
+                    await sleep(delay);
                     continue;
                 }
-                // Server error — retry once
-                if (response.status >= 500 && attempt === 0) {
-                    await sleep(2000);
+                // Server error — retry with backoff
+                if (response.status >= 500 && attempt < MAX_RETRIES - 1) {
+                    await sleep((attempt + 1) * 2000);
                     continue;
                 }
                 throw new Error(`LLM API error (${response.status}): ${errorText.slice(0, 500)}`);
             }
+            // Streaming path
+            if (useStream && response.body) {
+                return await readSSEStream(response.body, options.onChunk);
+            }
+            // Non-streaming path
             const json = await response.json();
             const choice = json.choices?.[0];
             if (!choice?.message?.content) {
@@ -96,9 +107,9 @@ export async function callLLM(prompt, config, systemPrompt) {
         }
         catch (err) {
             lastError = err;
-            // Network error — retry once
-            if (attempt === 0 && (err.code === 'ECONNREFUSED' || err.code === 'ETIMEDOUT' || err.message?.includes('fetch'))) {
-                await sleep(3000);
+            // Network error — retry with backoff
+            if (attempt < MAX_RETRIES - 1 && (err.code === 'ECONNREFUSED' || err.code === 'ETIMEDOUT' || err.message?.includes('fetch'))) {
+                await sleep((attempt + 1) * 3000);
                 continue;
             }
             throw err;
@@ -106,6 +117,46 @@ export async function callLLM(prompt, config, systemPrompt) {
     }
     throw lastError || new Error('LLM call failed after retries');
 }
+/**
+ * Read an SSE stream from an OpenAI-compatible streaming response.
+ */
+async function readSSEStream(body, onChunk) {
+    const decoder = new TextDecoder();
+    const reader = body.getReader();
+    let content = '';
+    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) {
+            const trimmed = line.trim();
+            if (!trimmed || !trimmed.startsWith('data: '))
+                continue;
+            const data = trimmed.slice(6);
+            if (data === '[DONE]')
+                continue;
+            try {
+                const parsed = JSON.parse(data);
+                const delta = parsed.choices?.[0]?.delta?.content;
+                if (delta) {
+                    content += delta;
+                    onChunk(content.length);
+                }
+            }
+            catch {
+                // Skip malformed SSE chunks
+            }
+        }
+    }
+    if (!content) {
+        throw new Error('LLM returned empty streaming response');
+    }
+    return { content };
+}
 function sleep(ms) {
     return new Promise(resolve => setTimeout(resolve, ms));
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "gitnexus",
-  "version": "1.2.3",
+  "version": "1.2.5",
   "description": "Graph-powered code intelligence for AI agents. Index any codebase, query via MCP or CLI.",
   "author": "Abhigyan Patwari",
   "license": "PolyForm-Noncommercial-1.0.0",