@crowdlisten/harness 1.0.0

package/dist/agent-tools.js

@@ -0,0 +1,409 @@
+/**
+ * Agent-Proxied Tools — Tools that proxy to agent.crowdlisten.com
+ *
+ * 5 skill packs, 16 tools. Each tool calls the agent backend via
+ * the shared agent-proxy helpers.
+ *
+ * Free vs Paid:
+ *   - Free: llm (2 tools), agent-network (2 tools) — no API key needed
+ *   - Paid: analysis (5), content (4), generation (2) — require CROWDLISTEN_API_KEY
+ */
+import { agentPost, agentGet, agentDelete, agentStream, requireApiKey, } from "./agent-proxy.js";
+// ─── Tool Definitions ──────────────────────────────────────────────────────
+export const AGENT_TOOLS = [
+    // ── Analysis Pack (5 tools) ────────────────────────────────────────────
+    {
+        name: "run_analysis",
+        description: "[Analysis] Run an audience analysis on a project. Streams results via SSE — returns the final analysis with themes, sentiment, insights, and recommendations. Requires CROWDLISTEN_API_KEY.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                project_id: { type: "string", description: "Project UUID" },
+                question: {
+                    type: "string",
+                    description: "Research question to analyze (e.g., 'What do users think about our pricing?')",
+                },
+                platforms: {
+                    type: "array",
+                    items: { type: "string" },
+                    description: "Platforms to search: reddit, youtube, tiktok, twitter, instagram, xiaohongshu. Default: all.",
+                },
+                max_results: {
+                    type: "number",
+                    description: "Max results per platform (default 20)",
+                },
+            },
+            required: ["project_id", "question"],
+        },
+    },
+    {
+        name: "continue_analysis",
+        description: "[Analysis] Continue a previous analysis with a follow-up question. Builds on existing context. Requires CROWDLISTEN_API_KEY.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                analysis_id: { type: "string", description: "Analysis UUID from run_analysis" },
+                question: { type: "string", description: "Follow-up question" },
+            },
+            required: ["analysis_id", "question"],
+        },
+    },
+    {
+        name: "get_analysis",
+        description: "[Analysis] Get the full results of a completed analysis including themes, sentiment breakdown, and source quotes. Requires CROWDLISTEN_API_KEY.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                analysis_id: { type: "string", description: "Analysis UUID" },
+            },
+            required: ["analysis_id"],
+        },
+    },
+    {
+        name: "list_analyses",
+        description: "[Analysis] List analyses for a project, newest first. Requires CROWDLISTEN_API_KEY.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                project_id: { type: "string", description: "Project UUID" },
+                limit: { type: "number", description: "Max results (default 20)" },
+            },
+            required: ["project_id"],
+        },
+    },
+    {
+        name: "generate_specs",
+        description: "[Analysis] Generate product specifications from analysis results — feature requests, user stories, and acceptance criteria extracted from audience signal. Requires CROWDLISTEN_API_KEY.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                project_id: { type: "string", description: "Project UUID" },
+                analysis_id: {
+                    type: "string",
+                    description: "Optional: scope to a specific analysis",
+                },
+                spec_type: {
+                    type: "string",
+                    description: "Type: feature_requests, user_stories, acceptance_criteria, all. Default: all.",
+                },
+            },
+            required: ["project_id"],
+        },
+    },
+    // ── Content Pack (4 tools) ─────────────────────────────────────────────
+    {
+        name: "ingest_content",
+        description: "[Content] Ingest content into the vector store for semantic search. Accepts raw text, URLs, or structured content. Requires CROWDLISTEN_API_KEY.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                project_id: { type: "string", description: "Project UUID" },
+                content: { type: "string", description: "Text content to ingest" },
+                source_url: { type: "string", description: "Optional: source URL" },
+                title: { type: "string", description: "Optional: content title" },
+                metadata: {
+                    type: "object",
+                    description: "Optional: key-value metadata",
+                },
+            },
+            required: ["project_id", "content"],
+        },
+    },
+    {
+        name: "search_vectors",
+        description: "[Content] Semantic search across ingested content using vector embeddings. Returns ranked results with relevance scores. Requires CROWDLISTEN_API_KEY.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                project_id: { type: "string", description: "Project UUID" },
+                query: { type: "string", description: "Search query" },
+                limit: { type: "number", description: "Max results (default 10)" },
+                threshold: {
+                    type: "number",
+                    description: "Minimum similarity score 0-1 (default 0.5)",
+                },
+            },
+            required: ["project_id", "query"],
+        },
+    },
+    {
+        name: "get_content_stats",
+        description: "[Content] Get statistics about ingested content — document count, total chunks, storage usage. Requires CROWDLISTEN_API_KEY.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                project_id: { type: "string", description: "Project UUID" },
+            },
+            required: ["project_id"],
+        },
+    },
+    {
+        name: "delete_content",
+        description: "[Content] Delete a specific content document and its vector embeddings. Requires CROWDLISTEN_API_KEY.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                content_id: { type: "string", description: "Content document UUID" },
+            },
+            required: ["content_id"],
+        },
+    },
+    // ── Generation Pack (2 tools) ──────────────────────────────────────────
+    {
+        name: "generate_prd",
+        description: "[Generation] Generate a Product Requirements Document from analysis results. Streams the PRD as it's generated. Requires CROWDLISTEN_API_KEY.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                project_id: { type: "string", description: "Project UUID" },
+                analysis_ids: {
+                    type: "array",
+                    items: { type: "string" },
+                    description: "Analysis UUIDs to base the PRD on",
+                },
+                template: {
+                    type: "string",
+                    description: "PRD template: standard, lean, technical, marketing. Default: standard.",
+                },
+                sections: {
+                    type: "array",
+                    items: { type: "string" },
+                    description: "Specific sections to generate (default: all). Options: overview, problem, solution, requirements, metrics, timeline",
+                },
+            },
+            required: ["project_id"],
+        },
+    },
+    {
+        name: "update_prd_section",
+        description: "[Generation] Update a specific section of an existing PRD with new content or instructions. Requires CROWDLISTEN_API_KEY.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                document_id: { type: "string", description: "PRD document UUID" },
+                section: {
+                    type: "string",
+                    description: "Section to update: overview, problem, solution, requirements, metrics, timeline",
+                },
+                instructions: {
+                    type: "string",
+                    description: "Instructions for how to update the section",
+                },
+                content: {
+                    type: "string",
+                    description: "Optional: replacement content (overrides instructions-based generation)",
+                },
+            },
+            required: ["document_id", "section"],
+        },
+    },
+    // ── LLM Pack (2 tools) — FREE ─────────────────────────────────────────
+    {
+        name: "llm_complete",
+        description: "[LLM] Send a completion request through CrowdListen's LLM proxy. Free tier — no API key required. Supports multiple models.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                prompt: { type: "string", description: "The prompt to complete" },
+                model: {
+                    type: "string",
+                    description: "Model to use (default: gpt-4o-mini). Call list_llm_models for available options.",
+                },
+                max_tokens: {
+                    type: "number",
+                    description: "Max tokens in response (default 1000)",
+                },
+                temperature: {
+                    type: "number",
+                    description: "Temperature 0-2 (default 0.7)",
+                },
+                system: {
+                    type: "string",
+                    description: "Optional system prompt",
+                },
+            },
+            required: ["prompt"],
+        },
+    },
+    {
+        name: "list_llm_models",
+        description: "[LLM] List available LLM models and their capabilities. Free — no API key required.",
+        inputSchema: { type: "object", properties: {} },
+    },
+    // ── Agent Network Pack (3 tools) — Mixed auth ─────────────────────────
+    {
+        name: "register_agent",
+        description: "[Agent Network] Register this agent in the CrowdListen agent network. Free — no API key required. Returns agent_id for future interactions.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                name: { type: "string", description: "Agent display name" },
+                capabilities: {
+                    type: "array",
+                    items: { type: "string" },
+                    description: "Agent capabilities: analysis, planning, coding, research, content",
+                },
+                executor: {
+                    type: "string",
+                    description: "Agent runtime: CLAUDE_CODE, CURSOR, GEMINI, CODEX, AMP, OPENCLAW",
+                },
+            },
+            required: ["name"],
+        },
+    },
+    {
+        name: "get_capabilities",
+        description: "[Agent Network] List capabilities of agents in the network. Free — no API key required.",
+        inputSchema: { type: "object", properties: {} },
+    },
+    {
+        name: "submit_analysis",
+        description: "[Agent Network] Submit an analysis result from this agent to the network for cross-agent synthesis. Requires CROWDLISTEN_API_KEY.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                agent_id: { type: "string", description: "Your agent_id from register_agent" },
+                analysis_id: { type: "string", description: "Analysis UUID to share" },
+                summary: { type: "string", description: "Brief summary of findings" },
+            },
+            required: ["agent_id", "analysis_id", "summary"],
+        },
+    },
+];
+// ─── Agent Tool Names (for dispatch routing) ───────────────────────────────
+const AGENT_TOOL_NAMES = new Set(AGENT_TOOLS.map((t) => t.name));
+export function isAgentTool(name) {
+    return AGENT_TOOL_NAMES.has(name);
+}
+// ─── Free tools (no API key required) ──────────────────────────────────────
+const FREE_TOOLS = new Set([
+    "llm_complete",
+    "list_llm_models",
+    "register_agent",
+    "get_capabilities",
+]);
+// ─── Handler ───────────────────────────────────────────────────────────────
+export async function handleAgentTool(name, args) {
+    const needsKey = !FREE_TOOLS.has(name);
+    const apiKey = needsKey ? requireApiKey() : process.env.CROWDLISTEN_API_KEY;
+    switch (name) {
+        // ── Analysis ───────────────────────────────────────────────
+        case "run_analysis": {
+            const result = await agentStream("/agent/v1/analysis/run", {
+                project_id: args.project_id,
+                question: args.question,
+                platforms: args.platforms,
+                max_results: args.max_results,
+            }, apiKey, 180_000 // 3 min timeout for analysis
+            );
+            return JSON.stringify(result, null, 2);
+        }
+        case "continue_analysis": {
+            const result = await agentPost(`/agent/v1/analysis/${args.analysis_id}/continue`, { question: args.question }, apiKey);
+            return JSON.stringify(result, null, 2);
+        }
+        case "get_analysis": {
+            const result = await agentGet(`/agent/v1/analysis/${args.analysis_id}`, apiKey);
+            return JSON.stringify(result, null, 2);
+        }
+        case "list_analyses": {
+            const limit = args.limit || 20;
+            const result = await agentGet(`/agent/v1/projects/${args.project_id}/analyses?limit=${limit}`, apiKey);
+            return JSON.stringify(result, null, 2);
+        }
+        case "generate_specs": {
+            const result = await agentPost("/agent/v1/specs/generate", {
+                project_id: args.project_id,
+                analysis_id: args.analysis_id,
+                spec_type: args.spec_type,
+            }, apiKey);
+            return JSON.stringify(result, null, 2);
+        }
+        // ── Content ────────────────────────────────────────────────
+        case "ingest_content": {
+            const result = await agentPost("/agent/v1/content/ingest", {
+                project_id: args.project_id,
+                content: args.content,
+                source_url: args.source_url,
+                title: args.title,
+                metadata: args.metadata,
+            }, apiKey);
+            return JSON.stringify(result, null, 2);
+        }
+        case "search_vectors": {
+            const result = await agentPost("/agent/v1/content/search", {
+                project_id: args.project_id,
+                query: args.query,
+                limit: args.limit,
+                threshold: args.threshold,
+            }, apiKey);
+            return JSON.stringify(result, null, 2);
+        }
+        case "get_content_stats": {
+            const result = await agentGet(`/agent/v1/content/stats?project_id=${args.project_id}`, apiKey);
+            return JSON.stringify(result, null, 2);
+        }
+        case "delete_content": {
+            const result = await agentDelete(`/agent/v1/content/${args.content_id}`, apiKey);
+            return JSON.stringify(result, null, 2);
+        }
+        // ── Generation ─────────────────────────────────────────────
+        case "generate_prd": {
+            const result = await agentStream("/agent/v1/prd/generate", {
+                project_id: args.project_id,
+                analysis_ids: args.analysis_ids,
+                template: args.template,
+                sections: args.sections,
+            }, apiKey, 180_000);
+            return JSON.stringify(result, null, 2);
+        }
+        case "update_prd_section": {
+            const result = await agentPost("/agent/v1/prd/update-section", {
+                document_id: args.document_id,
+                section: args.section,
+                instructions: args.instructions,
+                content: args.content,
+            }, apiKey);
+            return JSON.stringify(result, null, 2);
+        }
+        // ── LLM (Free) ────────────────────────────────────────────
+        case "llm_complete": {
+            const result = await agentPost("/api/v1/llm/complete", {
+                prompt: args.prompt,
+                model: args.model,
+                max_tokens: args.max_tokens,
+                temperature: args.temperature,
+                system: args.system,
+            }, apiKey // Optional — may be undefined for free tier
+            );
+            return JSON.stringify(result, null, 2);
+        }
+        case "list_llm_models": {
+            const result = await agentGet("/api/v1/llm/models", apiKey);
+            return JSON.stringify(result, null, 2);
+        }
+        // ── Agent Network ──────────────────────────────────────────
+        case "register_agent": {
+            const result = await agentPost("/api/agents/register", {
+                name: args.name,
+                capabilities: args.capabilities,
+                executor: args.executor,
+            }, apiKey);
+            return JSON.stringify(result, null, 2);
+        }
+        case "get_capabilities": {
+            const result = await agentGet("/api/agents/capabilities", apiKey);
+            return JSON.stringify(result, null, 2);
+        }
+        case "submit_analysis": {
+            const result = await agentPost("/api/agents/analyze", {
+                agent_id: args.agent_id,
+                analysis_id: args.analysis_id,
+                summary: args.summary,
+            }, apiKey);
+            return JSON.stringify(result, null, 2);
+        }
+        default:
+            throw new Error(`Unknown agent tool: ${name}`);
+    }
+}

package/dist/context/api.d.ts

@@ -0,0 +1,5 @@
+/**
+ * REST API routes for the context extraction web UI.
+ */
+import * as http from "http";
+export declare function handleApiRequest(req: http.IncomingMessage, res: http.ServerResponse, url: URL): Promise<void>;

package/dist/context/api.js

@@ -0,0 +1,164 @@
+/**
+ * REST API routes for the context extraction web UI.
+ */
+import { runPipeline } from "./pipeline.js";
+import { getBlocks, deleteBlock, updateBlock, clearBlocks, loadConfig, saveConfig, getHistory, } from "./store.js";
+import { matchSkills, discoverSkills, searchSkills, getSkillCategories } from "./matcher.js";
+import { loadUserState, activatePack, deactivatePack } from "./user-state.js";
+function readBody(req) {
+    return new Promise((resolve, reject) => {
+        let data = "";
+        req.on("data", (chunk) => (data += chunk.toString()));
+        req.on("end", () => resolve(data));
+        req.on("error", reject);
+    });
+}
+function json(res, data, status = 200) {
+    res.writeHead(status, { "Content-Type": "application/json" });
+    res.end(JSON.stringify(data));
+}
+function error(res, message, status = 400) {
+    json(res, { error: message }, status);
+}
+export async function handleApiRequest(req, res, url) {
+    const method = req.method || "GET";
+    const pathname = url.pathname;
+    try {
+        // POST /api/process — Process text through the pipeline
+        if (pathname === "/api/process" && method === "POST") {
+            const body = JSON.parse(await readBody(req));
+            const { text, source, isChat } = body;
+            if (!text) {
+                return error(res, "Missing 'text' field");
+            }
+            const result = await runPipeline({
+                text,
+                source: source || "web-upload",
+                isChat: isChat !== false,
+            });
+            return json(res, result);
+        }
+        // GET /api/blocks — Get stored context blocks
+        if (pathname === "/api/blocks" && method === "GET") {
+            return json(res, { blocks: getBlocks() });
+        }
+        // DELETE /api/blocks/:index — Delete a block by index
+        if (pathname.startsWith("/api/blocks/") && method === "DELETE") {
+            const index = parseInt(pathname.split("/").pop() || "", 10);
+            if (isNaN(index))
+                return error(res, "Invalid block index");
+            deleteBlock(index);
+            return json(res, { ok: true });
+        }
+        // PUT /api/blocks/:index — Update a block by index
+        if (pathname.startsWith("/api/blocks/") && method === "PUT") {
+            const index = parseInt(pathname.split("/").pop() || "", 10);
+            if (isNaN(index))
+                return error(res, "Invalid block index");
+            const body = JSON.parse(await readBody(req));
+            updateBlock(index, body);
+            return json(res, { ok: true });
+        }
+        // DELETE /api/blocks — Clear all blocks
+        if (pathname === "/api/blocks" && method === "DELETE") {
+            clearBlocks();
+            return json(res, { ok: true });
+        }
+        // GET /api/skills — Get skill recommendations from stored blocks
+        if (pathname === "/api/skills" && method === "GET") {
+            const blocks = getBlocks();
+            const skills = await matchSkills(blocks);
+            return json(res, { skills });
+        }
+        // GET /api/skills/discover — Context-driven skill discovery
+        if (pathname === "/api/skills/discover" && method === "GET") {
+            const blocks = getBlocks();
+            const category = url.searchParams.get("category");
+            const tier = url.searchParams.get("tier");
+            const limit = parseInt(url.searchParams.get("limit") || "10", 10);
+            const skills = await discoverSkills(blocks, {
+                category: category || undefined,
+                tier: tier || undefined,
+                limit,
+            });
+            return json(res, { skills });
+        }
+        // GET /api/skills/search — Full-text skill search
+        if (pathname === "/api/skills/search" && method === "GET") {
+            const q = url.searchParams.get("q") || "";
+            const tier = url.searchParams.get("tier");
+            const category = url.searchParams.get("category");
+            const skills = searchSkills(q, {
+                tier: tier || undefined,
+                category: category || undefined,
+            });
+            return json(res, { skills });
+        }
+        // GET /api/skills/categories — List categories with counts
+        if (pathname === "/api/skills/categories" && method === "GET") {
+            const categories = getSkillCategories();
+            return json(res, { categories });
+        }
+        // GET /api/config — Get current config status (redacted)
+        if (pathname === "/api/config" && method === "GET") {
+            const config = loadConfig();
+            if (!config) {
+                return json(res, { configured: false });
+            }
+            return json(res, {
+                configured: true,
+                provider: config.provider,
+                model: config.model || "default",
+                hasApiKey: !!config.apiKey,
+            });
+        }
+        // POST /api/config — Update LLM config
+        if (pathname === "/api/config" && method === "POST") {
+            const body = JSON.parse(await readBody(req));
+            if (!body.provider)
+                return error(res, "Missing 'provider' field");
+            if (!body.apiKey) {
+                return error(res, "API key required");
+            }
+            saveConfig(body);
+            return json(res, { ok: true });
+        }
+        // GET /api/history — Get processing history
+        if (pathname === "/api/history" && method === "GET") {
+            return json(res, { history: getHistory() });
+        }
+        // GET /api/state — Get UserState (active packs, preferences)
+        if (pathname === "/api/state" && method === "GET") {
+            return json(res, loadUserState());
+        }
+        // POST /api/state/activate — Activate a skill pack
+        if (pathname === "/api/state/activate" && method === "POST") {
+            const body = JSON.parse(await readBody(req));
+            if (!body.pack_id)
+                return error(res, "Missing 'pack_id'");
+            const state = activatePack(body.pack_id);
+            return json(res, { ok: true, activePacks: state.activePacks });
+        }
+        // POST /api/state/deactivate — Deactivate a skill pack
+        if (pathname === "/api/state/deactivate" && method === "POST") {
+            const body = JSON.parse(await readBody(req));
+            if (!body.pack_id)
+                return error(res, "Missing 'pack_id'");
+            const state = deactivatePack(body.pack_id);
+            return json(res, { ok: true, activePacks: state.activePacks });
+        }
+        // GET /api/packs — List all available packs with metadata
+        if (pathname === "/api/packs" && method === "GET") {
+            // Import registry dynamically to avoid circular deps at module load
+            const { listPacks } = await import("../tools/registry.js");
+            const state = loadUserState();
+            return json(res, { packs: listPacks(state.activePacks) });
+        }
+        // 404 for unknown API routes
+        error(res, "Not found", 404);
+    }
+    catch (err) {
+        const message = err instanceof Error ? err.message : String(err);
+        error(res, message, 500);
+    }
+}

package/dist/context/cli.d.ts

@@ -0,0 +1,19 @@
+/**
+ * CLI subcommands for context extraction.
+ *
+ * npx @crowdlisten/harness context          # Launch web UI
+ * npx @crowdlisten/harness context <file>   # CLI-only processing
+ * npx @crowdlisten/harness setup            # Configure LLM provider (handled in index.ts)
+ */
+/**
+ * Interactive setup: configure LLM provider and API key.
+ */
+export declare function runSetupContext(): Promise<void>;
+/**
+ * Process a file via CLI (no web UI).
+ */
+export declare function runContextCLI(filePath: string): Promise<void>;
+/**
+ * Launch the web UI server.
+ */
+export declare function runContextWeb(): Promise<void>;