haive.agents.discovery.selection_strategies¶

Tool selection strategies for dynamic tool selection.

This module implements various strategies for selecting tools based on different criteria and approaches, providing flexibility in how tools are chosen for different contexts and use cases.

Classes¶

`AdaptiveSelectionStrategy`	Adaptive selection that learns from usage patterns.
`BaseSelectionStrategy`	Base class for tool selection strategies.
`CapabilityBasedStrategy`	Capability-based tool selection.
`ContextualSelectionStrategy`	Context-aware tool selection considering conversation history.
`EnsembleSelectionStrategy`	Ensemble strategy combining multiple selection approaches.
`LearningSelectionStrategy`	Selection strategy that learns from user feedback and tool performance.
`SemanticSelectionStrategy`	Semantic similarity-based tool selection.

Functions¶

create_selection_strategy(strategy_name, **kwargs)

Create a selection strategy by name.

Module Contents¶

class haive.agents.discovery.selection_strategies.AdaptiveSelectionStrategy(learning_rate=0.1)¶

Bases: BaseSelectionStrategy

Adaptive selection that learns from usage patterns.

Init .

Parameters:: learning_rate (float) – [TODO: Add description]

async select_tools(query, available_tools, context, max_tools=5)¶

Select tools using adaptive learning.

Parameters:

query (str)
available_tools (list[haive.agents.discovery.semantic_discovery.ComponentMetadata])
context (haive.agents.discovery.dynamic_tool_selector.ContextAwareState)
max_tools (int)

Return type:

haive.agents.discovery.dynamic_tool_selector.ToolSelectionResult

update_performance(tool_name, success)¶

Update tool performance based on execution results.

Parameters:

tool_name (str)
success (bool)

Return type:

None

class haive.agents.discovery.selection_strategies.BaseSelectionStrategy¶

Bases: abc.ABC

Base class for tool selection strategies.

abstractmethod select_tools(query, available_tools, context, max_tools=5)¶

Async:

Parameters:

query (str)
available_tools (list[haive.agents.discovery.semantic_discovery.ComponentMetadata])
context (haive.agents.discovery.dynamic_tool_selector.ContextAwareState)
max_tools (int)

Return type:

haive.agents.discovery.dynamic_tool_selector.ToolSelectionResult

Select tools based on strategy.

class haive.agents.discovery.selection_strategies.CapabilityBasedStrategy(capability_weights=None)¶

Bases: BaseSelectionStrategy

Capability-based tool selection.

Init .

Parameters:: capability_weights (dict[str, float] | None) – [TODO: Add description]

async select_tools(query, available_tools, context, max_tools=5)¶

Select tools based on capability matching.

Parameters:

query (str)
available_tools (list[haive.agents.discovery.semantic_discovery.ComponentMetadata])
context (haive.agents.discovery.dynamic_tool_selector.ContextAwareState)
max_tools (int)

Return type:

haive.agents.discovery.dynamic_tool_selector.ToolSelectionResult

class haive.agents.discovery.selection_strategies.ContextualSelectionStrategy(context_weight=0.3)¶

Bases: BaseSelectionStrategy

Context-aware tool selection considering conversation history.

Init .

Parameters:: context_weight (float) – [TODO: Add description]

async select_tools(query, available_tools, context, max_tools=5)¶

Select tools considering full context.

Parameters:

query (str)
available_tools (list[haive.agents.discovery.semantic_discovery.ComponentMetadata])
context (haive.agents.discovery.dynamic_tool_selector.ContextAwareState)
max_tools (int)

Return type:

haive.agents.discovery.dynamic_tool_selector.ToolSelectionResult

class haive.agents.discovery.selection_strategies.EnsembleSelectionStrategy(strategies=None)¶

Bases: BaseSelectionStrategy

Ensemble strategy combining multiple selection approaches.

Init .

Parameters:: strategies (list[BaseSelectionStrategy] | None) – [TODO: Add description]

async select_tools(query, available_tools, context, max_tools=5)¶

Select tools using ensemble of strategies.

Parameters:

query (str)
available_tools (list[haive.agents.discovery.semantic_discovery.ComponentMetadata])
context (haive.agents.discovery.dynamic_tool_selector.ContextAwareState)
max_tools (int)

Return type:

haive.agents.discovery.dynamic_tool_selector.ToolSelectionResult

class haive.agents.discovery.selection_strategies.LearningSelectionStrategy¶

Bases: BaseSelectionStrategy

Selection strategy that learns from user feedback and tool performance.

Init .

Returns:: Add return description]
Return type:: [TODO

add_feedback(tool_name, rating, context, feedback_data)¶

Add user feedback for learning.

Parameters:

tool_name (str)
rating (float)
context (str)
feedback_data (dict[str, Any])

Return type:

None

async select_tools(query, available_tools, context, max_tools=5)¶

Select tools using learned patterns and feedback.

Parameters:

query (str)
available_tools (list[haive.agents.discovery.semantic_discovery.ComponentMetadata])
context (haive.agents.discovery.dynamic_tool_selector.ContextAwareState)
max_tools (int)

Return type:

haive.agents.discovery.dynamic_tool_selector.ToolSelectionResult

class haive.agents.discovery.selection_strategies.SemanticSelectionStrategy(similarity_threshold=0.7)¶

Bases: BaseSelectionStrategy

Semantic similarity-based tool selection.

Init .

Parameters:: similarity_threshold (float) – [TODO: Add description]

async select_tools(query, available_tools, context, max_tools=5)¶

Select tools based on semantic similarity to query.

Parameters:

query (str)
available_tools (list[haive.agents.discovery.semantic_discovery.ComponentMetadata])
context (haive.agents.discovery.dynamic_tool_selector.ContextAwareState)
max_tools (int)

Return type:

haive.agents.discovery.dynamic_tool_selector.ToolSelectionResult

haive.agents.discovery.selection_strategies.create_selection_strategy(strategy_name, **kwargs)¶

Create a selection strategy by name.

Parameters:: strategy_name (str)
Return type:: BaseSelectionStrategy