Module azure_chatbot 

Structs

AISearchOutput

ArrayProperty

ChatbotUserContext

Context about the user and course for a chatbot interaction. Passed to tool implementations so they can access user-specific data.

ContentFilter

ContentFilterResult

ContentFilterResults

GuardedStream 🔒

Custom stream that encapsulates both the response stream and the cancellation guard. Makes sure that the guard is always dropped when the stream is dropped.

IncompleteReason

Response received from LLM API

IncompleteResponse

Incomplete response received from LLM API

JsonItem

LLMRequest

LLMRequestResponseFormatParam

MistralParams

NonThinkingParams

Reasoning

ReasoningOutput

RequestCancelledGuard 🔒

RequestTextOptions

Response

Response received from LLM API

ResponseError

Error object returned by the LLM API on a failed response. Fields are optional so any error shape deserializes rather than crashing the stream parser.

ResponseOutput

Streamed token of the response text

Schema

Defines LLM structured output shape and types

StreamEventError

StreamItem 🔒

ThinkingParams

Enums

ArrayItem

ChatbotChatStreamEvent

ContentFilterSource

InputItem

JSONType

LLMRequestParams

LLMToolChoice

MessagePhase

Phase determines how to react to the message. Commentary-phase can have e.g. a reasoning preamble, and final answer is self-explanatory.

OutputItem

ParsedResponseLine 🔒

ResponseStreamType

SchemaPropertyType

StreamEvent 🔒

For passing streamed events and data between streaming functions.

SummaryType

Constants

ALL_EXPECTED_EVENTS 🔒

These are the events we expect to receive from Azure API but will not handle. This list doesn’t include the events we explicitly handle.

CONTENT_FIELD_SEPARATOR

SEARCH_GROUNDING_INSTRUCTION 🔒

Appended to the system prompt when course-material search is enabled, to ground answers in retrieved course material.

Functions

answer_unfinished_tool_calls 🔒

check_error_should_terminate_stream 🔒

These types of ChatbotErrors shouldn’t be shown to the user and are likely created from an unrecoverable error in our code that should make the stream fail.

error_event_string_from_message 🔒

Creates a ChatbotChatStreamEvent::Error from the message and returns it in string form. Either message or error should be Some. If message is Some, use it as the StreamEvent message. Else, use ChatbotError’s message. error is either a ResponseError received from Azure and stored in ChatbotError, or ChatbotError’s error message, or None.

make_request_and_create_stream

Makes a request to Azure and returns the resulting stream.

parse_text_response 🔒

Streams and parses an LLM response from Azure that contains a text response. Consumes the lines (stream) from Azure, because the stream ends when a text response is finished. Returns a stream to be consumed in the caller.

parse_tool 🔒

Streams and parses a LLM response from Azure that contains function calls. Calls the functions and yields a Vec of function results to be sent to Azure. Consumes the lines (stream), because it ends when a custom function call is made. Returns a stream to be consumed in the caller.

process_output_item

For saving output items that are not text messages or function calls, i.e. that don’t need further processing and are not streamed to the user. Saves reasoning and Azure AI Search items.

send_chat_request_and_parse_stream

Send and parse a Chatbot message and response and stream it to the user. Controls the whole operation.

stream_and_detect_response_stream_type 🔒

Stream from Azure and return the stream when a text response or tool call response is detected. Tool calls and text responses are processed later with differing logic. Returns a stream to be consumed in the caller. Yields the lines (stream) argument, which is the Azure stream.

Type Aliases

PeekableLinesStream 🔒

A LinesStream that is peekable. Needed to determine which type of LLM response is being received.