Package com.tnsai.feedback

Interface FeedbackCollector

public interface FeedbackCollector

SPI for collecting and learning from tool execution feedback.

When discovered via FeedbackCollector.Factory.discover(), the agent automatically:

• Records tool execution outcomes (success/failure/latency)
• Accepts explicit user feedback (corrections, preferences)
• Provides learned preferences for prompt augmentation

If tnsai-intelligence is not on the classpath, the agent operates without feedback collection (graceful degradation).

Implementation is provided by tnsai-intelligence module.

Since:

0.1.0

See Also:

• FeedbackCollector.ToolOutcome

Nested Class Summary

Nested Classes

Modifier and Type	Interface	Description
static interface	FeedbackCollector.Factory	Factory for creating FeedbackCollector instances.
static final record	FeedbackCollector.ToolOutcome	Immutable record of a tool execution outcome.

Method Summary

Modifier and Type	Method	Description
int	getFeedbackCount()	Returns the number of recorded feedback items.
List<String>	getLearnedPreferences()	Returns learned preferences from accumulated feedback.
List<String>	getPromptAdjustments()	Returns prompt adjustments derived from corrections and negative feedback.
static FeedbackCollector	noOp()	Returns a no-op implementation that silently ignores all operations.
void	recordToolOutcome(FeedbackCollector.ToolOutcome outcome)	Records the outcome of a tool execution.
void	recordUserFeedback(String type, String content, Map<String,Object> metadata)	Records explicit user feedback (thumbs up/down, correction, preference).

Method Details

noOp

static FeedbackCollector noOp()

Returns a no-op implementation that silently ignores all operations. Used as a null-object when tnsai-intelligence is not on the classpath.

Returns:

a no-op FeedbackCollector (singleton)

recordToolOutcome

void recordToolOutcome(FeedbackCollector.ToolOutcome outcome)

Records the outcome of a tool execution. Called automatically by the agent after each tool call.

Parameters:

outcome - the tool execution outcome

recordUserFeedback

void recordUserFeedback(String type,
 String content,
 Map<String,Object> metadata)

Records explicit user feedback (thumbs up/down, correction, preference).

Parameters:

type - feedback type (POSITIVE, NEGATIVE, CORRECTION, PREFERENCE)

content - the feedback content

metadata - optional metadata (tool name, session id, etc.)

getLearnedPreferences

List<String> getLearnedPreferences()

Returns learned preferences from accumulated feedback. Used by the agent to augment the system prompt.

Returns:

list of preference strings (never null, may be empty)

getPromptAdjustments

List<String> getPromptAdjustments()

Returns prompt adjustments derived from corrections and negative feedback. Used by the agent to augment the system prompt.

Returns:

list of prompt adjustment rules (never null, may be empty)

getFeedbackCount

int getFeedbackCount()

Returns the number of recorded feedback items.

Returns:

feedback count