Voice Call Analyzer with DeepAgents + TrueFoundry

Analyze voice call transcripts with 3 AI sub-agents running in parallel — each on a different LLM, all routed through one TrueFoundry AI Gateway. Action items are auto-created as Linear issues via TrueFoundry MCP Gateway.

🚀 Sign up for TrueFoundry  ·  📖 AI Gateway docs  ·  🤖 DeepAgents

---

Architecture

              load_transcript()
                     │
            Planner Agent (Claude Sonnet 4.6)
                     │
        ┌────────────┼────────────┐
        │            │            │
   Sentiment    Action Items     Coach
  (Gemini 3    (Claude 4.6 +   (GPT-5
   Flash)       Linear MCP)     Mini)
        │            │            │
        └────────────┼────────────┘
                     │
              Combined Report

3 LLMs, 1 API endpoint — TFY AI Gateway handles model routing, auth, rate limiting, and observability.

---

Quick Start

git clone <repo-url> && cd tfy-voice-analyser-agent
uv sync
cp .env.example .env

Fill in your credentials in .env:

# TFY AI Gateway (required)
TFY_GATEWAY_URL=https://gateway.truefoundry.ai
TFY_API_KEY=tfy-...

# Linear project name (optional)
LINEAR_PROJECT=

# TFY MCP Gateway for Linear (optional — skips issue creation if not set)
TFY_MCP_GATEWAY_URL=https://gateway.truefoundry.ai/your-org/mcp/linear/server
TFY_MCP_GATEWAY_KEY=tfy-...

Start the server:

langgraph dev --port 8888

This gives you:

• API: http://localhost:8888
• Studio: https://smith.langchain.com/studio/?baseUrl=http://localhost:8888
• Docs: http://localhost:8888/docs

---

What It Does

1. Loads a sample support call transcript (~8 min call)
2. Spawns 3 sub-agents in parallel, each on a different LLM via Gateway:
   • Sentiment Analyzer (Gemini Flash) — tone, emotional arc, CSAT score
   • Action Items Creator (Claude Sonnet) — extracts items + creates Linear issues via MCP
   • Call Coach (GPT-5 Mini) — strengths, improvements, suggested phrases
3. Combines outputs into a single report
4. Returns the report with links to created Linear issues

---

Customizing Models

Model names in agent.py must match what's registered on your TFY AI Gateway. Update these to match your setup:

llm("flash/gemini-3-flash")                        # sentiment
llm("bedrock/global.anthropic.claude-sonnet-4-6")   # action items + planner
llm("openai-main/gpt-5-mini")                       # coaching

---

Files

agent.py              — the whole thing (~130 lines)
langgraph.json        — graph config for langgraph dev
sample_transcript.txt — sample support call for demo
pyproject.toml        — dependencies
.env.example          — credential template

---

What This Demonstrates

Component	What	Why
DeepAgents	Multi-agent orchestration with parallel sub-agents	Planning, delegation, combining results
TFY AI Gateway	Single endpoint routing to 3 different LLMs	Model routing, cost control, observability
TFY MCP Gateway	Linear issue creation via MCP protocol	Secure tool access with auth + audit