Uploaded May 2026 | Updated September 2026, 2 weeks ago
In this clip from Bill's Ultimate AI Workshop, he dives into the core mechanics of implementing AI tool calling "step two" of building a custom AI chat agent. Using a simple weather-checking tool, we explore how to configure your model to request external tools.
Bill covers the exact steps needed to manually maintain your context window during chat completions, including how to capture the model's requested tool call and properly append the tool ID, function name, and arguments back into the conversation history. We also look at how to return structured JSON data to the model under the "tool" role so it can confidently answer the user's original prompt.
What you'll learn in this clip:
• Phase 2: Setting up tool calling mechanics and parameters
• Handling potential multi-tool calls using slices
• The critical role of Tool IDs and updating the conversation history
• Feeding structured data responses back into the context window
• Why simple tool parameters prevent models from hallucinating or failing
A crucial best practice for AI tool design Bill discusses: why you should build small, "Linux-like" tools with simple parameters to ensure that even lower-level models can reason effectively without making errors.
---
Explore more from Ardan Labs
Online Courses: ardanlabs.com/education
Live Training Events: ardanlabs.com/live-training-events
Technical Blog: ardanlabs.com/blog
Github: github.com/ardanlabs
---
Connect with Ardan Labs
Website: ardanlabs.com
X: https://x.com/ardanlabs
LinkedIn: linkedin.com/company/ardanlabs
Kronk AI: kronkai.com
#AIChatAgent #golang #llm #sse #machinelearning
In this clip from Bill's Ultimate AI Workshop, he dives into the core mechanics of implementing AI tool calling "step two" of building a custom AI chat agent. Using a simple weather-checking tool, we explore how to configure your model to request external tools.
Bill covers the exact steps needed to manually maintain your context window during chat completions, including how to capture the model's requested tool call and properly append the tool ID, function name, and arguments back into the conversation history. We also look at how to return structured JSON data to the model under the "tool" role so it can confidently answer the user's original prompt.
What you'll learn in this clip:
• Phase 2: Setting up tool calling mechanics and parameters
• Handling potential multi-tool calls using slices
• The critical role of Tool IDs and updating the conversation history
• Feeding structured data responses back into the context window
• Why simple tool parameters prevent models from hallucinating or failing
A crucial best practice for AI tool design Bill discusses: why you should build small, "Linux-like" tools with simple parameters to ensure that even lower-level models can reason effectively without making errors.
---
Explore more from Ardan Labs
Online Courses: ardanlabs.com/education
Live Training Events: ardanlabs.com/live-training-events
Technical Blog: ardanlabs.com/blog
Github: github.com/ardanlabs
---
Connect with Ardan Labs
Website: ardanlabs.com
X: https://x.com/ardanlabs
LinkedIn: linkedin.com/company/ardanlabs
Kronk AI: kronkai.com
#AIChatAgent #golang #llm #sse #machinelearning
