How-to Guides¶

Task recipes. Each page solves one concrete problem and shows a minimal working example in C++, Python, and Unreal where applicable. If you are new to Tryll, start with First inference; if you want the why, see Concepts.

Running the server¶

Run the Tryll Server — install or build from source, edit server-config.json, start, watch logs.

Sessions and agents¶

Connect and Manage a Session — connect, configure, reconnect, clean shutdown.
Build a Chat Agent with a Graph — compose a Generate + guardrail graph.
Stream Answers to a UI — wire streaming tokens into your application's view layer.

Models¶

Use Your Own Local Model — register a GGUF file that lives on disk without a Hugging Face download.
Pin and Unpin Models — keep a model warm (LoadModelRequest) or release it (UnloadModelRequest).
Enable Auto Model Downloading — let CreateAgent download missing models automatically during development; eliminates the explicit download-then-create loop.

Retrieval-augmented generation¶

Create a Simple RAG Assistant ★ — the hero recipe: prepare a knowledge base, add a Retrieve node in front of Generate, see grounded answers.
Use Mustache Templates — control exactly how retrieved chunks and instructions land in the prompt.

Guardrails and scripted branches¶

Use Canned Responses and Guardrails — short-circuit jailbreaks and off-topic prompts to scripted replies.

Voice input¶

Use Voice Input — create a VoiceInput handle, stream PCM audio, receive transcripts, push-to-talk vs. hands-free.
Bias Voice Input with Hotwords — nudge the STT decoder toward game-specific proper nouns, spell names, and other lexicon items without retraining the model.

Tools¶

Define and Handle Tool Calls — declare ToolDefs, route on tool_called / no_tool_called, and handle OnToolCall / set_on_tool_call / SetOnToolCall client-side.

Concepts — the mental model behind each recipe.
Reference — field-level detail for every knob.