Build Advanced Chatbots with Semantic Kernel

This is a set of 11 Semantic Kernel Python concept samples covering chat completion patterns beyond a basic chatbot: a simple chatbot baseline (simple_chatbot.py), calling it as a kernel function (simple_chatbot_kernel_function.py), applying logit bias (simple_chatbot_logit_bias.py), storing metadata (simple_chatbot_store_metadata.py), streaming responses (simple_chatbot_streaming.py), handling image input (simple_chatbot_with_image.py), and five variations of conversation history management: summary-based history reduction (plain, auto-reducing, and content-preserving variants) and truncation-based history reduction (plain and auto-reducing variants).

Use these as reference implementations for the specific chat completion capability you need: streaming for responsive UIs, logit bias for constraining token generation, image input for multimodal chat, or - notably - one of the five history-management samples when a conversation is growing too long for the model's context window and needs either summarization or truncation to stay within limits.

Not a single all-in-one chatbot template - each file demonstrates one capability in isolation, so pick the sample matching your specific need (e.g. summary-reducer vs truncation-reducer) rather than expecting one file to cover everything.

Inputs: user chat messages (text or, for the image sample, image input), plus growing conversation history for the history-reducer samples.

Outputs: chat completion responses - streamed or complete - with conversation history managed according to the chosen reduction strategy (summarized or truncated, with or without automatic triggering).

Microsoft Semantic Kernel (Python), covering its chat completion, streaming, and ChatHistoryReducer (summary and truncation variants) abstractions.

Developers building Semantic Kernel chatbots who need a specific capability - streaming, logit bias, image input, or long-conversation history management - as working reference code.