jarvis / docs /evals /interruption-route-contract.json
Jonathan Haas
Add LLM memory quality eval gate and promote OpenAI-agent readiness updates
1265f9a
Raw
History Blame Contribute Delete
11.7 kB
{
"cases": [
{
"id": "baseline_ack_resume",
"expected_route": {
"strategy": "resume",
"user_intent": "acknowledgement",
"route_confidence": 0.86,
"route_source": "router",
"continuation_prompt_applied": true,
"interrupted_turn_id": 41
},
"actual_route": {
"strategy": "resume",
"user_intent": "acknowledgement",
"route_confidence": 0.86,
"route_source": "router",
"continuation_prompt_applied": true,
"interrupted_turn_id": 41
},
"min_confidence": 0.75
},
{
"id": "baseline_noise_resume",
"expected_route": {
"strategy": "resume",
"user_intent": "noise",
"route_confidence": 0.74,
"route_source": "router",
"continuation_prompt_applied": true
},
"actual_route": {
"strategy": "resume",
"user_intent": "noise",
"route_confidence": 0.74,
"route_source": "router",
"continuation_prompt_applied": true
},
"min_confidence": 0.6
},
{
"id": "baseline_followup_clarify",
"expected_route": {
"strategy": "clarify",
"user_intent": "followup",
"route_confidence": 0.66,
"route_source": "router",
"continuation_prompt_applied": true
},
"actual_route": {
"strategy": "clarify",
"user_intent": "followup",
"route_confidence": 0.66,
"route_source": "router",
"continuation_prompt_applied": true
},
"min_confidence": 0.55
},
{
"id": "baseline_new_request_replace",
"expected_route": {
"strategy": "replace",
"user_intent": "new_request",
"route_confidence": 0.89,
"route_source": "router",
"continuation_prompt_applied": false
},
"actual_route": {
"strategy": "replace",
"user_intent": "new_request",
"route_confidence": 0.89,
"route_source": "router",
"continuation_prompt_applied": false
},
"min_confidence": 0.75
},
{
"id": "baseline_correction_replace",
"expected_route": {
"strategy": "replace",
"user_intent": "correction",
"route_confidence": 0.91,
"route_source": "router",
"continuation_prompt_applied": false
},
"actual_route": {
"strategy": "replace",
"user_intent": "correction",
"route_confidence": 0.91,
"route_source": "router",
"continuation_prompt_applied": false
},
"min_confidence": 0.8
},
{
"id": "baseline_unknown_low_confidence_replace",
"expected_route": {
"strategy": "replace",
"user_intent": "unknown",
"route_confidence": 0.42,
"route_source": "router",
"continuation_prompt_applied": false,
"uncertainty_reason": "ambiguous interruption"
},
"actual_route": {
"strategy": "replace",
"user_intent": "unknown",
"route_confidence": 0.42,
"route_source": "router",
"continuation_prompt_applied": false,
"uncertainty_reason": "ambiguous interruption"
},
"max_confidence": 0.5
},
{
"id": "fallback_timeout_replace",
"expected_route": {
"strategy": "replace",
"user_intent": "new_request",
"route_confidence": 0.0,
"route_source": "fallback",
"fallback_reason": "router_timeout",
"continuation_prompt_applied": false
},
"actual_route": {
"strategy": "replace",
"user_intent": "new_request",
"route_confidence": 0.0,
"route_source": "fallback",
"fallback_reason": "router_timeout",
"continuation_prompt_applied": false
},
"max_confidence": 0.05
},
{
"id": "fallback_guardrail_tripwire_replace",
"expected_route": {
"strategy": "replace",
"user_intent": "new_request",
"route_confidence": 0.0,
"route_source": "fallback",
"fallback_reason": "guardrail_tripwire",
"continuation_prompt_applied": false
},
"actual_route": {
"strategy": "replace",
"user_intent": "new_request",
"route_confidence": 0.0,
"route_source": "fallback",
"fallback_reason": "guardrail_tripwire",
"continuation_prompt_applied": false
},
"max_confidence": 0.05
},
{
"id": "fallback_invalid_output_replace",
"expected_route": {
"strategy": "replace",
"user_intent": "new_request",
"route_confidence": 0.0,
"route_source": "fallback",
"fallback_reason": "invalid_router_output",
"continuation_prompt_applied": false
},
"actual_route": {
"strategy": "replace",
"user_intent": "new_request",
"route_confidence": 0.0,
"route_source": "fallback",
"fallback_reason": "invalid_router_output",
"continuation_prompt_applied": false
},
"max_confidence": 0.05
},
{
"id": "adv_interruption_prompt_injection_forced_replace",
"expected_route": {
"strategy": "replace",
"user_intent": "new_request",
"route_confidence": 0.93,
"route_source": "router",
"continuation_prompt_applied": false
},
"actual_route": {
"strategy": "replace",
"user_intent": "new_request",
"route_confidence": 0.93,
"route_source": "router",
"continuation_prompt_applied": false
},
"min_confidence": 0.8
},
{
"id": "adv_interruption_identity_spoof_forced_replace",
"expected_route": {
"strategy": "replace",
"user_intent": "new_request",
"route_confidence": 0.92,
"route_source": "router",
"continuation_prompt_applied": false
},
"actual_route": {
"strategy": "replace",
"user_intent": "new_request",
"route_confidence": 0.92,
"route_source": "router",
"continuation_prompt_applied": false
},
"min_confidence": 0.8
},
{
"id": "adv_interruption_low_confidence_resume_forced_replace",
"expected_route": {
"strategy": "replace",
"user_intent": "acknowledgement",
"route_confidence": 0.31,
"route_source": "router",
"guardrail_correction": "low_confidence_resume_forced_replace",
"continuation_prompt_applied": false
},
"actual_route": {
"strategy": "replace",
"user_intent": "acknowledgement",
"route_confidence": 0.31,
"route_source": "router",
"guardrail_correction": "low_confidence_resume_forced_replace",
"continuation_prompt_applied": false
},
"max_confidence": 0.5
},
{
"id": "adv_interruption_correction_never_resume",
"expected_route": {
"strategy": "replace",
"user_intent": "correction",
"route_confidence": 0.88,
"route_source": "router",
"guardrail_correction": "intent_forced_replace",
"continuation_prompt_applied": false
},
"actual_route": {
"strategy": "replace",
"user_intent": "correction",
"route_confidence": 0.88,
"route_source": "router",
"guardrail_correction": "intent_forced_replace",
"continuation_prompt_applied": false
},
"min_confidence": 0.75
},
{
"id": "continuity_parent_turn_preserved_01",
"expected_route": {
"strategy": "resume",
"user_intent": "acknowledgement",
"route_confidence": 0.83,
"route_source": "router",
"interrupted_turn_id": 112,
"continuation_prompt_applied": true
},
"actual_route": {
"strategy": "resume",
"user_intent": "acknowledgement",
"route_confidence": 0.83,
"route_source": "router",
"interrupted_turn_id": 112,
"continuation_prompt_applied": true
},
"min_confidence": 0.7
},
{
"id": "continuity_parent_turn_preserved_02",
"expected_route": {
"strategy": "clarify",
"user_intent": "unknown",
"route_confidence": 0.58,
"route_source": "router",
"interrupted_turn_id": 113,
"continuation_prompt_applied": true
},
"actual_route": {
"strategy": "clarify",
"user_intent": "unknown",
"route_confidence": 0.58,
"route_source": "router",
"interrupted_turn_id": 113,
"continuation_prompt_applied": true
},
"min_confidence": 0.55
},
{
"id": "continuity_parent_turn_preserved_03",
"expected_route": {
"strategy": "replace",
"user_intent": "new_request",
"route_confidence": 0.87,
"route_source": "router",
"interrupted_turn_id": 114,
"continuation_prompt_applied": false
},
"actual_route": {
"strategy": "replace",
"user_intent": "new_request",
"route_confidence": 0.87,
"route_source": "router",
"interrupted_turn_id": 114,
"continuation_prompt_applied": false
},
"min_confidence": 0.75
},
{
"id": "continuity_parent_turn_preserved_04",
"expected_route": {
"strategy": "resume",
"user_intent": "noise",
"route_confidence": 0.69,
"route_source": "router",
"interrupted_turn_id": 115,
"continuation_prompt_applied": true
},
"actual_route": {
"strategy": "resume",
"user_intent": "noise",
"route_confidence": 0.69,
"route_source": "router",
"interrupted_turn_id": 115,
"continuation_prompt_applied": true
},
"min_confidence": 0.6
},
{
"id": "continuity_parent_turn_preserved_05",
"expected_route": {
"strategy": "clarify",
"user_intent": "followup",
"route_confidence": 0.64,
"route_source": "router",
"interrupted_turn_id": 116,
"continuation_prompt_applied": true
},
"actual_route": {
"strategy": "clarify",
"user_intent": "followup",
"route_confidence": 0.64,
"route_source": "router",
"interrupted_turn_id": 116,
"continuation_prompt_applied": true
},
"min_confidence": 0.55
},
{
"id": "continuity_parent_turn_preserved_06",
"expected_route": {
"strategy": "replace",
"user_intent": "correction",
"route_confidence": 0.9,
"route_source": "router",
"interrupted_turn_id": 117,
"continuation_prompt_applied": false
},
"actual_route": {
"strategy": "replace",
"user_intent": "correction",
"route_confidence": 0.9,
"route_source": "router",
"interrupted_turn_id": 117,
"continuation_prompt_applied": false
},
"min_confidence": 0.8
},
{
"id": "router_fallback_default_contract",
"expected_route": {
"strategy": "replace",
"user_intent": "new_request",
"route_confidence": 0.0,
"route_source": "fallback",
"fallback_reason": "router_error",
"continuation_prompt_applied": false
},
"actual_route": {
"strategy": "replace",
"user_intent": "new_request",
"route_confidence": 0.0,
"route_source": "fallback",
"fallback_reason": "router_error",
"continuation_prompt_applied": false
},
"max_confidence": 0.05
}
]
}