RetrievalConfig class
Retrieval and context budget constants.
These are provider-agnostic — they control how content chunks are retrieved and how conversation context is managed, regardless of which AI provider is active.
Properties
- hashCode → int
-
The hash code for this object.
no setterinherited
- runtimeType → Type
-
A representation of the runtime type of the object.
no setterinherited
Methods
-
noSuchMethod(
Invocation invocation) → dynamic -
Invoked when a nonexistent method or property is accessed.
inherited
-
toString(
) → String -
A string representation of this object.
inherited
Operators
-
operator ==(
Object other) → bool -
The equality operator.
inherited
Constants
- ftsNeighborWindow → const int
- Number of neighboring chunks to include around each search hit.
- ftsTopK → const int
- Default max chunks returned by FTS search.
- maxChunkChars → const int
- Context budget allocation (in characters)
- maxChunkCharsLocal → const int
- Reference-material budget for the on-device llama LLM. nCtx=4096 tokens leaves ~3-4k chars after the pedagogical preamble + mode prompt + ctx block + conversation; anything more silently truncates the system prompt and the topic anchor is the first thing the model loses sight of.
- maxConversationChars → const int
-
modeTopK
→ const Map<
String, int> - Per-mode top-K overrides for search-based retrieval.
-
questionDrivenModes
→ const Set<
String> - Chat modes that use search-based retrieval (FTS or semantic).
- semanticThreshold → const double
- Default similarity threshold for semantic search.
-
sequentialModes
→ const Set<
String> - Chat modes that load all chunks in document order.