trustgraph/docs/tech-specs/agent-explainability.md

273 lines
9.8 KiB
Markdown
Raw Normal View History

Adding explainability to the ReACT agent (#689) * Added tech spec * Add provenance recording to React agent loop Enables agent sessions to be traced and debugged using the same explainability infrastructure as GraphRAG. Agent traces record: - Session start with query and timestamp - Each iteration's thought, action, arguments, and observation - Final answer with derivation chain Changes: - Add session_id and collection fields to AgentRequest schema - Add agent predicates (TG_THOUGHT, TG_ACTION, etc.) to namespaces - Create agent provenance triple generators in provenance/agent.py - Register explainability producer in agent service - Emit provenance triples during agent execution - Update CLI tools to detect and render agent traces alongside GraphRAG * Updated explainability taxonomy: GraphRAG: tg:Question → tg:Exploration → tg:Focus → tg:Synthesis Agent: tg:Question → tg:Analysis(s) → tg:Conclusion All entities also have their PROV-O type (prov:Activity or prov:Entity). Updated commit message: Add provenance recording to React agent loop Enables agent sessions to be traced and debugged using the same explainability infrastructure as GraphRAG. Entity types follow human reasoning patterns: - tg:Question - the user's query (shared with GraphRAG) - tg:Analysis - each think/act/observe cycle - tg:Conclusion - the final answer Also adds explicit TG types to GraphRAG entities: - tg:Question, tg:Exploration, tg:Focus, tg:Synthesis All types retain their PROV-O base types (prov:Activity, prov:Entity). Changes: - Add session_id and collection fields to AgentRequest schema - Add explainability entity types to namespaces.py - Create agent provenance triple generators - Register explainability producer in agent service - Emit provenance triples during agent execution - Update CLI tools to detect and render both trace types * Document RAG explainability is now complete. Here's a summary of the changes made: Schema Changes: - trustgraph-base/trustgraph/schema/services/retrieval.py: Added explain_id and explain_graph fields to DocumentRagResponse - trustgraph-base/trustgraph/messaging/translators/retrieval.py: Updated translator to handle explainability fields Provenance Changes: - trustgraph-base/trustgraph/provenance/namespaces.py: Added TG_CHUNK_COUNT and TG_SELECTED_CHUNK predicates - trustgraph-base/trustgraph/provenance/uris.py: Added docrag_question_uri, docrag_exploration_uri, docrag_synthesis_uri generators - trustgraph-base/trustgraph/provenance/triples.py: Added docrag_question_triples, docrag_exploration_triples, docrag_synthesis_triples builders - trustgraph-base/trustgraph/provenance/__init__.py: Exported all new Document RAG functions and predicates Service Changes: - trustgraph-flow/trustgraph/retrieval/document_rag/document_rag.py: Added explainability callback support and triple emission at each phase (Question → Exploration → Synthesis) - trustgraph-flow/trustgraph/retrieval/document_rag/rag.py: Registered explainability producer and wired up the callback Documentation: - docs/tech-specs/agent-explainability.md: Added Document RAG entity types and provenance model documentation Document RAG Provenance Model: Question (urn:trustgraph:docrag:{uuid}) │ │ tg:query, prov:startedAtTime │ rdf:type = prov:Activity, tg:Question │ ↓ prov:wasGeneratedBy │ Exploration (urn:trustgraph:docrag:{uuid}/exploration) │ │ tg:chunkCount, tg:selectedChunk (multiple) │ rdf:type = prov:Entity, tg:Exploration │ ↓ prov:wasDerivedFrom │ Synthesis (urn:trustgraph:docrag:{uuid}/synthesis) │ │ tg:content = "The answer..." │ rdf:type = prov:Entity, tg:Synthesis * Specific subtype that makes the retrieval mechanism immediately obvious: System: GraphRAG TG Types on Question: tg:Question, tg:GraphRagQuestion URI Pattern: urn:trustgraph:question:{uuid} ──────────────────────────────────────── System: Document RAG TG Types on Question: tg:Question, tg:DocRagQuestion URI Pattern: urn:trustgraph:docrag:{uuid} ──────────────────────────────────────── System: Agent TG Types on Question: tg:Question, tg:AgentQuestion URI Pattern: urn:trustgraph:agent:{uuid} Files modified: - trustgraph-base/trustgraph/provenance/namespaces.py - Added TG_GRAPH_RAG_QUESTION, TG_DOC_RAG_QUESTION, TG_AGENT_QUESTION - trustgraph-base/trustgraph/provenance/triples.py - Added subtype to question_triples and docrag_question_triples - trustgraph-base/trustgraph/provenance/agent.py - Added subtype to agent_session_triples - trustgraph-base/trustgraph/provenance/__init__.py - Exported new types - docs/tech-specs/agent-explainability.md - Documented the subtypes This allows: - Query all questions: ?q rdf:type tg:Question - Query only GraphRAG: ?q rdf:type tg:GraphRagQuestion - Query only Document RAG: ?q rdf:type tg:DocRagQuestion - Query only Agent: ?q rdf:type tg:AgentQuestion * Fixed tests
2026-03-11 15:28:15 +00:00
# Agent Explainability: Provenance Recording
## Overview
Add provenance recording to the React agent loop so agent sessions can be traced and debugged using the same explainability infrastructure as GraphRAG.
**Design Decisions:**
- Write to `urn:graph:retrieval` (generic explainability graph)
- Linear dependency chain for now (analysis N → wasDerivedFrom → analysis N-1)
- Tools are opaque black boxes (record input/output only)
- DAG support deferred to future iteration
## Entity Types
Both GraphRAG and Agent use PROV-O as the base ontology with TrustGraph-specific subtypes:
### GraphRAG Types
| Entity | PROV-O Type | TG Types | Description |
|--------|-------------|----------|-------------|
| Question | `prov:Activity` | `tg:Question`, `tg:GraphRagQuestion` | The user's query |
| Exploration | `prov:Entity` | `tg:Exploration` | Edges retrieved from knowledge graph |
| Focus | `prov:Entity` | `tg:Focus` | Selected edges with reasoning |
| Synthesis | `prov:Entity` | `tg:Synthesis` | Final answer |
### Agent Types
| Entity | PROV-O Type | TG Types | Description |
|--------|-------------|----------|-------------|
| Question | `prov:Activity` | `tg:Question`, `tg:AgentQuestion` | The user's query |
| Analysis | `prov:Entity` | `tg:Analysis` | Each think/act/observe cycle |
| Conclusion | `prov:Entity` | `tg:Conclusion` | Final answer |
### Document RAG Types
| Entity | PROV-O Type | TG Types | Description |
|--------|-------------|----------|-------------|
| Question | `prov:Activity` | `tg:Question`, `tg:DocRagQuestion` | The user's query |
| Exploration | `prov:Entity` | `tg:Exploration` | Chunks retrieved from document store |
| Synthesis | `prov:Entity` | `tg:Synthesis` | Final answer |
**Note:** Document RAG uses a subset of GraphRAG's types (no Focus step since there's no edge selection/reasoning phase).
### Question Subtypes
All Question entities share `tg:Question` as a base type but have a specific subtype to identify the retrieval mechanism:
| Subtype | URI Pattern | Mechanism |
|---------|-------------|-----------|
| `tg:GraphRagQuestion` | `urn:trustgraph:question:{uuid}` | Knowledge graph RAG |
| `tg:DocRagQuestion` | `urn:trustgraph:docrag:{uuid}` | Document/chunk RAG |
| `tg:AgentQuestion` | `urn:trustgraph:agent:{uuid}` | ReAct agent |
This allows querying all questions via `tg:Question` while filtering by specific mechanism via the subtype.
## Provenance Model
```
Question (urn:trustgraph:agent:{uuid})
│ tg:query = "User's question"
│ prov:startedAtTime = timestamp
│ rdf:type = prov:Activity, tg:Question
↓ prov:wasDerivedFrom
Analysis1 (urn:trustgraph:agent:{uuid}/i1)
│ tg:thought = "I need to query the knowledge base..."
│ tg:action = "knowledge-query"
│ tg:arguments = {"question": "..."}
│ tg:observation = "Result from tool..."
│ rdf:type = prov:Entity, tg:Analysis
↓ prov:wasDerivedFrom
Analysis2 (urn:trustgraph:agent:{uuid}/i2)
│ ...
↓ prov:wasDerivedFrom
Conclusion (urn:trustgraph:agent:{uuid}/final)
│ tg:answer = "The final response..."
│ rdf:type = prov:Entity, tg:Conclusion
```
### Document RAG Provenance Model
```
Question (urn:trustgraph:docrag:{uuid})
│ tg:query = "User's question"
│ prov:startedAtTime = timestamp
│ rdf:type = prov:Activity, tg:Question
↓ prov:wasGeneratedBy
Exploration (urn:trustgraph:docrag:{uuid}/exploration)
│ tg:chunkCount = 5
│ tg:selectedChunk = "chunk-id-1"
│ tg:selectedChunk = "chunk-id-2"
│ ...
│ rdf:type = prov:Entity, tg:Exploration
↓ prov:wasDerivedFrom
Synthesis (urn:trustgraph:docrag:{uuid}/synthesis)
│ tg:content = "The synthesized answer..."
│ rdf:type = prov:Entity, tg:Synthesis
```
## Changes Required
### 1. Schema Changes
**File:** `trustgraph-base/trustgraph/schema/services/agent.py`
Add `session_id` and `collection` fields to `AgentRequest`:
```python
@dataclass
class AgentRequest:
question: str = ""
state: str = ""
group: list[str] | None = None
history: list[AgentStep] = field(default_factory=list)
user: str = ""
collection: str = "default" # NEW: Collection for provenance traces
streaming: bool = False
session_id: str = "" # NEW: For provenance tracking across iterations
```
**File:** `trustgraph-base/trustgraph/messaging/translators/agent.py`
Update translator to handle `session_id` and `collection` in both `to_pulsar()` and `from_pulsar()`.
### 2. Add Explainability Producer to Agent Service
**File:** `trustgraph-flow/trustgraph/agent/react/service.py`
Register an "explainability" producer (same pattern as GraphRAG):
```python
from ... base import ProducerSpec
from ... schema import Triples
# In __init__:
self.register_specification(
ProducerSpec(
name = "explainability",
schema = Triples,
)
)
```
### 3. Provenance Triple Generation
**File:** `trustgraph-base/trustgraph/provenance/agent.py`
Create helper functions (similar to GraphRAG's `question_triples`, `exploration_triples`, etc.):
```python
def agent_session_triples(session_uri, query, timestamp):
"""Generate triples for agent Question."""
return [
Triple(s=session_uri, p=RDF_TYPE, o=PROV_ACTIVITY),
Triple(s=session_uri, p=RDF_TYPE, o=TG_QUESTION),
Triple(s=session_uri, p=TG_QUERY, o=query),
Triple(s=session_uri, p=PROV_STARTED_AT_TIME, o=timestamp),
]
def agent_iteration_triples(iteration_uri, parent_uri, thought, action, arguments, observation):
"""Generate triples for one Analysis step."""
return [
Triple(s=iteration_uri, p=RDF_TYPE, o=PROV_ENTITY),
Triple(s=iteration_uri, p=RDF_TYPE, o=TG_ANALYSIS),
Triple(s=iteration_uri, p=TG_THOUGHT, o=thought),
Triple(s=iteration_uri, p=TG_ACTION, o=action),
Triple(s=iteration_uri, p=TG_ARGUMENTS, o=json.dumps(arguments)),
Triple(s=iteration_uri, p=TG_OBSERVATION, o=observation),
Triple(s=iteration_uri, p=PROV_WAS_DERIVED_FROM, o=parent_uri),
]
def agent_final_triples(final_uri, parent_uri, answer):
"""Generate triples for Conclusion."""
return [
Triple(s=final_uri, p=RDF_TYPE, o=PROV_ENTITY),
Triple(s=final_uri, p=RDF_TYPE, o=TG_CONCLUSION),
Triple(s=final_uri, p=TG_ANSWER, o=answer),
Triple(s=final_uri, p=PROV_WAS_DERIVED_FROM, o=parent_uri),
]
```
### 4. Type Definitions
**File:** `trustgraph-base/trustgraph/provenance/namespaces.py`
Add explainability entity types and agent predicates:
```python
# Explainability entity types (used by both GraphRAG and Agent)
TG_QUESTION = TG + "Question"
TG_EXPLORATION = TG + "Exploration"
TG_FOCUS = TG + "Focus"
TG_SYNTHESIS = TG + "Synthesis"
TG_ANALYSIS = TG + "Analysis"
TG_CONCLUSION = TG + "Conclusion"
# Agent predicates
TG_THOUGHT = TG + "thought"
TG_ACTION = TG + "action"
TG_ARGUMENTS = TG + "arguments"
TG_OBSERVATION = TG + "observation"
TG_ANSWER = TG + "answer"
```
## Files Modified
| File | Change |
|------|--------|
| `trustgraph-base/trustgraph/schema/services/agent.py` | Add session_id and collection to AgentRequest |
| `trustgraph-base/trustgraph/messaging/translators/agent.py` | Update translator for new fields |
| `trustgraph-base/trustgraph/provenance/namespaces.py` | Add entity types, agent predicates, and Document RAG predicates |
| `trustgraph-base/trustgraph/provenance/triples.py` | Add TG types to GraphRAG triple builders, add Document RAG triple builders |
| `trustgraph-base/trustgraph/provenance/uris.py` | Add Document RAG URI generators |
| `trustgraph-base/trustgraph/provenance/__init__.py` | Export new types, predicates, and Document RAG functions |
| `trustgraph-base/trustgraph/schema/services/retrieval.py` | Add explain_id and explain_graph to DocumentRagResponse |
| `trustgraph-base/trustgraph/messaging/translators/retrieval.py` | Update DocumentRagResponseTranslator for explainability fields |
| `trustgraph-flow/trustgraph/agent/react/service.py` | Add explainability producer + recording logic |
| `trustgraph-flow/trustgraph/retrieval/document_rag/document_rag.py` | Add explainability callback and emit provenance triples |
| `trustgraph-flow/trustgraph/retrieval/document_rag/rag.py` | Add explainability producer and wire up callback |
| `trustgraph-cli/trustgraph/cli/show_explain_trace.py` | Handle agent trace types |
| `trustgraph-cli/trustgraph/cli/list_explain_traces.py` | List agent sessions alongside GraphRAG |
## Files Created
| File | Purpose |
|------|---------|
| `trustgraph-base/trustgraph/provenance/agent.py` | Agent-specific triple generators |
## CLI Updates
**Detection:** Both GraphRAG and Agent Questions have `tg:Question` type. Distinguished by:
1. URI pattern: `urn:trustgraph:agent:` vs `urn:trustgraph:question:`
2. Derived entities: `tg:Analysis` (agent) vs `tg:Exploration` (GraphRAG)
**`list_explain_traces.py`:**
- Shows Type column (Agent vs GraphRAG)
**`show_explain_trace.py`:**
- Auto-detects trace type
- Agent rendering shows: Question → Analysis step(s) → Conclusion
## Backwards Compatibility
- `session_id` defaults to `""` - old requests work, just won't have provenance
- `collection` defaults to `"default"` - reasonable fallback
- CLI gracefully handles both trace types
## Verification
```bash
# Run an agent query
tg-invoke-agent -q "What is the capital of France?"
# List traces (should show agent sessions with Type column)
tg-list-explain-traces -U trustgraph -C default
# Show agent trace
tg-show-explain-trace "urn:trustgraph:agent:xxx"
```
## Future Work (Not This PR)
- DAG dependencies (when analysis N uses results from multiple prior analyses)
- Tool-specific provenance linking (KnowledgeQuery → its GraphRAG trace)
- Streaming provenance emission (emit as we go, not batch at end)