Skip to main content

What You Can Do

The endpoint supports a wide range of capabilities to power intelligent applications:
  • Basic Chat → Natural conversation in Swahili, English, and other African languages.
  • Multimodal → Combine text and images as input for richer interactions.
  • RAG (Retrieval-Augmented Generation) → Connect external knowledge sources for contextual, grounded responses.
  • Agentic RAG → Extend RAG with agent workflows for planning, reasoning, and task execution.
  • Agents → Run multi-step autonomous workflows where the model coordinates actions.
  • Reasoning → Handle structured problem solving, logic, and advanced query resolution.
  • Tools Calling → Dynamically call external APIs or functions from within a conversation.
  • Structured Output → Generate reliable, typed JSON for downstream processing.

Chat Models includes

The chat endpoint runs on Pawa small language models (SLMs), optimized for efficiency, long context, and multilingual use.
  • Pawa Ember → Lightweight, efficient model for real-time and low-cost deployments.
  • Pawa Blaze → Advanced model for reasoning, multimodal tasks, and agentic workflows.

With chat endpoint, you can start simple and scale up to advanced multimodal and agent-based applications — without changing the core interface.
I