AI agents

// packages/ai/src/agents/search/index.ts
import { Output, ToolLoopAgent } from "ai";
import { claudeHaiku } from "../../models";
import { searchTools } from "./tools";
import { SEARCH_INSTRUCTIONS } from "./prompts";

function createInternalSearchAgent(
  context: SearchToolContext,
  hasTools: boolean,
) {
  return new ToolLoopAgent({
    model: claudeHaiku,
    tools: hasTools ? searchTools : undefined,
    instructions: SEARCH_INSTRUCTIONS,
    experimental_context: context,
    output: Output.object({ schema: searchOutputSchema }),
  });
}

import { agent } from "@feedback/ai/agents";

const result = await agent.search.generate({
  query: "Users want to export their data to CSV format",
});

// result.output.matches = [
//   {
//     requestId: "req_123",
//     title: "CSV Export Feature",
//     description: "Allow users to export data as CSV",
//     confidence: 0.92,
//     reason: "Exact match - both describe CSV export functionality"
//   },
//   ...
// ]

// packages/ai/src/agents/search/prompts.ts
export const SEARCH_INSTRUCTIONS = `
Confidence scoring:
- 0.9-1.0: Query describes the exact same feature/issue
- 0.8-0.89: Strong match, same general feature with minor differences
- 0.7-0.79: Moderate match, related but not identical
- 0.5-0.69: Weak match but potentially relevant
- Below 0.5: Not a match, don't include

Return all matches with confidence >= 0.5, ordered by highest confidence first.
`;

import { agent } from "@feedback/ai/agents";

// Create a new feature request
const result = await agent.request.generate({
  customerPain: "Users can't export their data",
  mode: "create",
});

// result.output = {
//   type: "create",
//   title: "Data Export Functionality",
//   description: "Users need the ability to export their data...",
//   areaIds: ["area_analytics", "area_data"],
// }

// Try to match, create if no match found
const result = await agent.request.generate({
  customerPain: "Need better CSV export options",
  mode: "match_or_create",
});

// result.output = {
//   type: "match_or_create",
//   decision: "matched", // or "created"
//   requestId: "req_123",
//   confidence: 0.85,
//   reason: "Matches existing CSV export request",
// }

import { agent } from "@feedback/ai/agents";

// Extract feedback from Slack thread
const result = await agent.slack.generate({
  mode: "extract",
  ref: { channel: "C123", threadTs: "1234567890.123456" },
});

// result.output = {
//   type: "extraction",
//   painDescription: "Customer can't export reports in CSV format",
//   isCustomerSpecific: true,
//   source: "thread",
// }

// Generate a notification message
const result = await agent.slack.generate({
  mode: "compose",
  composeType: "match_notification",
  composeContext: {
    requestUrl: "https://app.example.com/requests/csv-export",
    requestTitle: "CSV Export Feature",
    summary: "Added your feedback to the CSV export request",
  },
});

// result.output = {
//   type: "compose",
//   message: "✅ Added your feedback to CSV Export Feature\n...",
// }

import { createAreaInsightsAgent } from "@feedback/ai/agents/area-insights";

// Create agent with pre-fetched context
const agent = createAreaInsightsAgent({
  areaId: "area_analytics",
  slug: "analytics",
  requests: [...], // Open requests with feedback
  entries: [...],  // Feedback entries
  accounts: [...], // Account data
});

const result = await agent.generate({
  prompt: "Analyze the analytics product area",
});

// result.output = {
//   areaId: "area_analytics",
//   slug: "analytics",
//   generatedAt: "2026-03-04T10:30:00Z",
//   sections: [
//     {
//       id: "dealbreakers",
//       title: "Critical Dealbreakers",
//       summary: "3 requests blocking $2.5M in ARR",
//       importance: "critical",
//       visuals: [
//         {
//           kind: "dealbreakers-metric",
//           props: {
//             count: 3,
//             totalArr: 2500000,
//             topItems: [...],
//           },
//         },
//       ],
//     },
//     ...
//   ],
// }

// packages/ai/src/agents/area-insights.ts
const AREA_INSIGHTS_INSTRUCTIONS = `
You are a product analytics strategist analyzing OPEN feature requests.

Your job is to produce a concise AreaInsightsReport with up to 4 sections:

1. "dealbreakers" - Critical issues blocking revenue
2. "segments" - Which customer segments feel the most pain
3. "themes" - Recurring patterns of confusion or friction
4. "execution" - How well this area is executing

Guidelines:
- Be opinionated: highlight what matters most
- Keep summaries to 1-2 sentences max, no fluff
- Set importance to "critical" only for dealbreakers with >$500k ARR
- If a section has no meaningful data, skip it entirely
- Order sections by importance (critical first)
`;

import { agent } from "@feedback/ai/agents";

const result = await agent.identity.generate({
  email: "[email protected]",
});

// result.output = {
//   userId: "user_123",
//   confidence: 0.95,
//   reason: "Exact email match",
// }

// packages/ai/src/agents/search/tools.ts
import { tool } from "ai";
import { z } from "zod";

export const searchRequests = tool({
  description:
    "Search for request items matching a query. Uses semantic search if available.",
  inputSchema: z.object({
    query: z.string().describe("The search query"),
    limit: z.number().optional().default(50).describe("Max results to return"),
    excludeIds: z
      .array(z.string())
      .optional()
      .describe("IDs to exclude from results"),
  }),
  execute: async ({ query, limit, excludeIds }, { experimental_context }) => {
    const ctx = experimental_context as SearchToolContext;
    
    // Create embedding for semantic search
    const embedding = await ctx.createEmbedding(query);
    
    if (embedding) {
      // Perform vector similarity search
      const searchResults = await ctx.searchSimilar(embedding, limit, excludeIds);
      
      // Fetch full request details
      const ids = searchResults.map((r) => r.id);
      const details = await ctx.fetchRequestsByIds(ids);
      
      return details.map((d) => ({
        ...d,
        score: searchResults.find((r) => r.id === d.id)?.score ?? 0,
      }));
    }
    
    // Fallback to recent open requests
    return await db.query.requests.findMany({
      where: (requests, { eq }) => eq(requests.status, "open"),
      orderBy: (requests, { desc }) => [desc(requests.createdAt)],
      limit: limit ?? 50,
    });
  },
});

import { createAgent } from "@feedback/ai/agents";

const customAgent = createAgent({
  search: {
    createEmbedding: async (text) => {
      // Custom embedding logic
      return myEmbeddingService.embed(text);
    },
    searchSimilar: async (embedding, limit) => {
      // Custom vector search
      return myVectorDB.search(embedding, limit);
    },
  },
});

import { Output } from "ai";
import { z } from "zod";

const searchOutputSchema = z.object({
  matches: z
    .array(
      z.object({
        requestId: z.string(),
        title: z.string(),
        description: z.string(),
        confidence: z.number().min(0).max(1),
        reason: z.string().optional(),
      }),
    )
    .describe("Matches ordered by highest confidence first"),
});

const agent = new ToolLoopAgent({
  model: claudeHaiku,
  tools: searchTools,
  instructions: SEARCH_INSTRUCTIONS,
  output: Output.object({ schema: searchOutputSchema }),
});

import { tool } from "ai";
import { z } from "zod";

export const yourTool = tool({
  description: "What this tool does",
  inputSchema: z.object({
    param: z.string(),
  }),
  execute: async ({ param }, { experimental_context }) => {
    // Tool logic here
  },
});

export const yourTools = { yourTool };

export const YOUR_AGENT_INSTRUCTIONS = `
You are an AI agent that...

Your job is to...

Guidelines:
- Be specific
- Use tools to fetch data
- Return structured output
`;

import { ToolLoopAgent, Output } from "ai";
import { z } from "zod";
import { claudeHaiku } from "../../models";
import { yourTools } from "./tools";
import { YOUR_AGENT_INSTRUCTIONS } from "./prompts";

const outputSchema = z.object({
  result: z.string(),
});

export function createYourAgent() {
  return new ToolLoopAgent({
    model: claudeHaiku,
    tools: yourTools,
    instructions: YOUR_AGENT_INSTRUCTIONS,
    output: Output.object({ schema: outputSchema }),
  });
}

// packages/ai/src/agents/index.ts
import { createYourAgent } from "./your-agent";

export function createAgent(overrides = {}) {
  return {
    // ... existing agents
    yourAgent: createYourAgent(),
  };
}