fireflies-core-workflow-a

Fireflies.ai Core Workflow A -- Transcript Retrieval & Processing

Overview

Primary workflow for Fireflies.ai: fetch meeting transcripts via GraphQL, process speaker-diarized sentences, extract action items and summaries, and route meeting intelligence downstream.

Prerequisites

• Completed fireflies-install-auth setup
• FIREFLIES_API_KEY set with Business+ plan for full access
• At least one completed meeting in Fireflies

Instructions

Step 1: Build the GraphQL Client

// lib/fireflies.ts
const FIREFLIES_API = "https://api.fireflies.ai/graphql";

export async function firefliesQuery<T = any>(
  query: string,
  variables?: Record<string, any>
): Promise<T> {
  const res = await fetch(FIREFLIES_API, {
    method: "POST",
    headers: {
      "Content-Type": "application/json",
      Authorization: `Bearer ${process.env.FIREFLIES_API_KEY}`,
    },
    body: JSON.stringify({ query, variables }),
  });

  const json = await res.json();
  if (json.errors) {
    const err = json.errors[0];
    throw new Error(`Fireflies API error: ${err.message} (${err.code || "unknown"})`);
  }
  return json.data;
}

Step 2: List Transcripts with Filters

const LIST_TRANSCRIPTS = `
  query ListTranscripts(
    $limit: Int,
    $mine: Boolean,
    $fromDate: DateTime,
    $toDate: DateTime,
    $organizers: [String],
    $participants: [String]
  ) {
    transcripts(
      limit: $limit
      mine: $mine
      fromDate: $fromDate
      toDate: $toDate
      organizers: $organizers
      participants: $participants
    ) {
      id title date duration
      organizer_email participants
      summary { overview action_items keywords }
    }
  }
`;

// Fetch this week's meetings for a specific organizer
const data = await firefliesQuery(LIST_TRANSCRIPTS, {
  limit: 20,
  fromDate: "2026-03-15T00:00:00Z",
  organizers: ["[email protected]"],
});

Step 3: Fetch Full Transcript with Sentences

const GET_TRANSCRIPT = `
  query GetTranscript($id: String!) {
    transcript(id: $id) {
      id title date duration
      organizer_email
      speakers { id name }
      sentences {
        index
        speaker_name
        speaker_id
        text
        raw_text
        start_time
        end_time
        ai_filters {
          task
          question
          sentiment
          pricing
          metric
          date_and_time
        }
      }
      summary {
        overview
        short_summary
        bullet_gist
        action_items
        keywords
        outline
        topics_discussed
      }
      meeting_attendees { displayName email }
      meeting_attendance { name join_time leave_time }
      analytics {
        sentiments { positive_pct negative_pct neutral_pct }
        speakers {
          name duration word_count
          words_per_minute questions
          longest_monologue filler_words
        }
      }
    }
  }
`;

const { transcript } = await firefliesQuery(GET_TRANSCRIPT, { id: "abc123" });

Step 4: Process Meeting Intelligence

interface MeetingIntelligence {
  id: string;
  title: string;
  attendees: string[];
  actionItems: string[];
  keyTopics: string[];
  speakerBreakdown: { name: string; minutes: number; wordCount: number }[];
  sentiment: { positive: number; negative: number; neutral: number };
  questions: string[];
}

function processMeeting(transcript: any): MeetingIntelligence {
  // Extract questions from AI filters
  const questions = transcript.sentences
    .filter((s: any) => s.ai_filters?.question)
    .map((s: any) => `${s.speaker_name}: ${s.text}`);

  return {
    id: transcript.id,
    title: transcript.title,
    attendees: transcript.meeting_attendees?.map((a: any) => a.email) || [],
    actionItems: transcript.summary?.action_items || [],
    keyTopics: transcript.summary?.keywords || [],
    speakerBreakdown: (transcript.analytics?.speakers || []).map((s: any) => ({
      name: s.name,
      minutes: Math.round(s.duration / 60),
      wordCount: s.word_count,
    })),
    sentiment: {
      positive: transcript.analytics?.sentiments?.positive_pct || 0,
      negative: transcript.analytics?.sentiments?.negative_pct || 0,
      neutral: transcript.analytics?.sentiments?.neutral_pct || 0,
    },
    questions,
  };
}

Step 5: Export Transcript as Text

function transcriptToText(transcript: any): string {
  const lines: string[] = [
    `# ${transcript.title}`,
    `Date: ${transcript.date} | Duration: ${transcript.duration}min`,
    `Speakers: ${transcript.speakers.map((s: any) => s.name).join(", ")}`,
    "",
    "## Summary",
    transcript.summary?.overview || "(no summary)",
    "",
    "## Action Items",
    ...(transcript.summary?.action_items || []).map((a: string) => `- ${a}`),
    "",
    "## Transcript",
  ];

  for (const s of transcript.sentences) {
    const timestamp = formatTimestamp(s.start_time);
    lines.push(`[${timestamp}] ${s.speaker_name}: ${s.text}`);
  }
  return lines.join("\n");
}

function formatTimestamp(seconds: number): string {
  const m = Math.floor(seconds / 60);
  const s = Math.floor(seconds % 60);
  return `${String(m).padStart(2, "0")}:${String(s).padStart(2, "0")}`;
}

Error Handling

Output

• Full transcript with speaker-diarized sentences and timestamps
• AI-generated summary, action items, and keywords
• Speaker analytics with talk time, word count, and sentiment
• Meeting intelligence object ready for downstream processing

Resources

• Transcript Query
• Fireflies API Concepts

Next Steps

For search, analytics, and AskFred, see fireflies-core-workflow-b.