Skip to main content

Post Session Debrief

Comprehensive post-session debrief workflow demonstrating discovery and deep analysis.

Driver wants to analyze a Global Mazda MX-5 Cup session from September 2025 at Lime Rock Park. This scenario demonstrates the complete workflow from data discovery through detailed performance analysis:

  1. Discovery: Query by date/track to find candidate sessions
  2. Overview: High-level session summary and lap statistics
  3. Best Lap Focus: Deep dive into single best lap performance
  4. Consistency Analysis: Compare performance across all laps
  5. Telemetry Visualization: Detailed technique analysis with graphics
  6. Braking Analysis: Brake point consistency and technique assessment
  7. Theoretical Best: Calculate achievable lap time from best mini-sectors
  8. Track Heatmap: Visual speed map showing performance patterns

Uses known session (01K451BYJ64G0T2GPQ44AKXC86) for reproducible demonstration. Combines parallel analyzer execution, lap selection strategies, and visualization.

Demonstration includes 8 turns covering 7 of 12 available analyzers:

  • Data analyzers: sessions_overview, lapstats, cornering, braking, theoretical_best
  • Graphics analyzers: telemetry_overlay, track_heatmap

Scenario Details​

  • Turns: 8
  • Set: agent_scenarios

Conversation Flow​

1. Show me my recent sessions at Lime Rock​

Agent Thinking:

User is requesting sessions by location. I'll query using
the track parameter and a reasonable limit to get recent sessions.

Tool Calls:

  • analyze(source="garage61", track="Lime Rock", limit=100, format="markdown", analyzers=[sessions_overview])

Agent Response:

Here are your recent sessions at Lime Rock Park:

Global Mazda MX-5 Cup​
  • Total Events: 3
  • Total Laps: 16
  • Total Time on Track: 21m 9s

Session 1 (01K451BYJ64G0T2GPQ44AKXC86)

  • Date: 2025-09-02 11:08
  • Driver: Marcel Hild
  • Laps: 7 (3 clean)
  • Best Lap: 62.708s
  • Clean %: 42%

Session 2 (01K1AWSR94V62ETMN67EGBMSP4)

  • Date: 2025-07-29 10:57
  • Driver: Marcel Hild
  • Laps: 7 (4 clean)
  • Best Lap: 62.800s

Insights:

  • Your best lap time was 62.708 seconds in session 01K451BYJ64G0T2GPQ44AKXC86
  • You have room to improve clean lap percentage

2. Give me detailed lap stats for that first MX-5 session​

Note: This turn requires context from previous turns.

Agent Thinking:

User identified the specific MX-5 session from the previous list
and wants detailed lap time statistics. I already showed them the high-level
session overview in the previous turn, so now I'll focus on lap-by-lap
statistics including best lap, consistency, and lap time distribution.

Tool Calls:

  • analyze(source="garage61", event="01K451BYJ64G0T2GPQ44AKXC86", format="markdown", analyzers=[lapstats])

Agent Response:

Here are the detailed lap statistics for your MX-5 session at Lime Rock Park (01K451BYJ64G0T2GPQ44AKXC86):

Session Details:

  • Track: Lime Rock Park
  • Car: Global Mazda MX-5 Cup
  • Driver: Marcel Hild
  • Session ID: 01K451BYJ64G0T2GPQ44AKXC86

Lap Statistics​

  • Best Lap: 62.708s
  • Worst Lap: 64.778s
  • Average Lap Time: 63.4s
  • Lap Count: 5 valid laps (2 outliers filtered)
  • Consistency: 1.18% coefficient of variation

Your most consistent laps were in the 62-63 second range.

3. Let's analyze my best lap corner by corner. Where am I lo...​

Note: This turn requires context from previous turns.

Agent Thinking:

User wants detailed corner-by-corner breakdown of their best lap to identify
where they're losing time. I'll analyze the best lap using the cornering
analyzer with track_order sorting so corners appear sequentially around
the track. This makes it easier to follow lap progression and identify
specific problem areas.

Tool Calls:

  • analyze(source="garage61", event="01K451BYJ64G0T2GPQ44AKXC86", format="markdown", analyzers=[cornering], cornering-lap-selection="best", cornering-sort-order="track_order")

Agent Response:

Here's the corner-by-corner analysis of your best lap in the MX-5 session at Lime Rock Park (01K451BYJ64G0T2GPQ44AKXC86):

Overall Performance Summary​

  • Average Entry Speed: 35.5 km/h
  • Average Apex Speed: 30.3 km/h
  • Average Exit Speed: 32.2 km/h
  • Average Corner Time: 5.4 seconds

Key Findings by Corner​

Turn 1 (Big Bend):

  • Entry Speed: 27-44 km/h (highly variable)
  • Exit Speed: 28-33 km/h
  • Insight: Significant entry speed variation suggests inconsistent braking points

Turn 2:

  • Entry Speed: 28-34 km/h
  • Exit Speed: 29-31 km/h
  • Best Corner Time: 3.75s
  • Insight: Exit speed varies significantly, indicating inconsistent throttle application

Turn 3:

  • Entry: 28-30 km/h (consistent)
  • Exit: ~32 km/h
  • Insight: Most consistent corner

Turn 4:

  • Entry Speed: 20-39 km/h (very inconsistent)
  • Exit Speed: 20-25 km/h
  • Insight: Major speed drop indicates technique issues

Turn 5:

  • Entry: ~33 km/h
  • Exit: 35-37 km/h
  • Insight: Good performance with higher exit speeds

Turn 6:

  • Entry: 38-39 km/h
  • Exit: 39-41 km/h
  • Insight: Well-executed, maintain this approach

Recommendations​

  • Focus on consistency in Turns 1 and 2 where speed variations are highest
  • Review braking points for smoother corner transitions
  • Work on steady throttle application through corners

If you want to dive deeper into your consistency across all laps or explore other aspects of this session (01K451BYJ64G0T2GPQ44AKXC86), let me know!

4. How consistent am I across all my laps in that session?​

Note: This turn requires context from previous turns.

Agent Thinking:

User wants to understand consistency across all laps, not just the best.
I'll run cornering with lap-selection: all to show variation in corner
performance across the entire session. The analyzer now includes lap count,
consistency percentages (coefficient of variation), and filters out bad
data (laps with <10 km/h speeds). This reveals which corners are consistent
vs inconsistent, and how many valid laps were analyzed.

Tool Calls:

  • analyze(source="garage61", event="01K451BYJ64G0T2GPQ44AKXC86", format="markdown", analyzers=[cornering], cornering-lap-selection="all")

5. Show me telemetry for Big Bend - that's my trouble spot​

Note: This turn requires context from previous turns.

Agent Thinking:

User wants to visualize telemetry for Big Bend (Turn 1) to diagnose
their technique issues in that corner. I'll filter telemetry to just
that corner section and show brake point markers to help understand
their braking and throttle application through the corner.

Tool Calls:

  • analyze(source="garage61", event="01K451BYJ64G0T2GPQ44AKXC86", lap_selection="best", format="markdown", analyzers=[telemetry_overlay], telemetry_overlay-corner-filter="big_bend", telemetry_overlay-show-brake-points=True)

6. Analyze my braking across all corners​

Note: This turn requires context from previous turns.

Agent Thinking:

User wants a comprehensive braking analysis. I'll run the braking
analyzer to show brake points, pressure, consistency, and trail
braking metrics across all corners. This complements the corner
telemetry view by providing overall braking statistics.

Tool Calls:

  • analyze(source="garage61", event="01K451BYJ64G0T2GPQ44AKXC86", format="markdown", analyzers=[braking])

Agent Response:

Here's your braking analysis:

Braking Performance Summary​

Overall Metrics:

  • Brake zones detected: 6 per lap
  • Average brake pressure: 82%
  • Brake consistency score: 78%
  • Trail braking score: 0.45/1.0

Key Findings:

Brake Pressure: You're averaging 82% of maximum brake pressure. You have more braking performance available - the car can handle 95-100%.

Consistency: Your brake points vary by 10-15m between laps. The fastest laps show later, more committed braking. The slower laps show earlier, more tentative braking.

Trail Braking: Score of 0.45 indicates you're releasing brakes too abruptly. More gradual release (trail braking) would help rotation and maintain corner speed.

Recommendation: Pick consistent brake markers and commit to them. Your best laps prove you can brake later - make that your default, not your exception.

7. What's my theoretical best lap if I put all my best secto...​

Note: This turn requires context from previous turns.

Agent Thinking:

User wants to understand their ultimate potential by combining best
mini-sector times from different laps. This shows how much time is
"on the table" with their current skill level and proves they can
go faster by putting together their best segments.

Tool Calls:

  • analyze(source="garage61", event="01K451BYJ64G0T2GPQ44AKXC86", format="markdown", analyzers=[theoretical_best])

8. Show me where on track I'm fastest and slowest​

Note: This turn requires context from previous turns.

Agent Thinking:

User wants a visual summary showing speed patterns across the track.
A speed heatmap provides intuitive visualization of performance,
making it easy to identify problem areas. I'll use the best lap
to show their optimal racing line with speed colored from slow (purple)
to fast (yellow) in km/h for easy interpretation.

Tool Calls:

  • analyze(source="garage61", event="01K451BYJ64G0T2GPQ44AKXC86", format="markdown", analyzers=[track_heatmap], track_heatmap-lap-selection="best", track_heatmap-value-column="speed")