| # | Role | Agent ID | Public IP | Private IP |
|---|---|---|---|---|
| 1 | agent | ai:alice | 104.236.92.237 | 10.11.0.3 |
| 2 | agent | ai:bob | 104.236.54.83 | 10.11.0.2 |
| 3 | agent | ai:charlie | 104.236.96.109 | 10.11.0.4 |
| 4 | memory-only | — | 104.131.108.255 | 10.11.0.5 |
Per the authoritative baseline spec, every agent node must emit a self-attestation before any scenario is permitted to run. This run's attestation:
Spec version: 1.4.0 — see authoritative baseline.
| Node | Agent | Framework | Authentic | MCP ai-memory | xAI cfg | xAI default | Agent ID | Federation | UFW off | iptables | dead-man | F1 xAI | F2a substrate | F2b agent (non-gating) | Config SHA | Pass |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| node-1 | ai:alice | hermes Hermes Agent v0.10.0 (2026.4.16) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | — | fa358f9a9059 | PASS |
| node-2 | ai:bob | hermes Hermes Agent v0.10.0 (2026.4.16) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | — | 21635cf63640 | PASS |
| node-3 | ai:charlie | hermes Hermes Agent v0.10.0 (2026.4.16) | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | — | ce52d772ef5a | PASS |
{
"baseline_pass": true,
"per_node": [
{
"spec_version": "1.4.0",
"agent_type": "hermes",
"agent_id": "ai:alice",
"node_index": "1",
"framework_version": "Hermes Agent v0.10.0 (2026.4.16)",
"ai_memory_version": "v0.6.2",
"peer_urls": "http://10.11.0.2:9077,http://10.11.0.4:9077,http://10.11.0.5:9077",
"config_file_sha256": "fa358f9a90597243fb96224babd541399bd7b1e972f364605308ab1e2d9dd2c7",
"config_attestation": {
"framework_is_authentic": true,
"mcp_server_ai_memory_registered": true,
"llm_backend_is_xai_grok": true,
"llm_is_default_provider": true,
"mcp_command_is_ai_memory": true,
"agent_id_stamped": true,
"federation_live": true,
"ufw_disabled": true,
"iptables_flushed": true,
"dead_man_switch_scheduled": true
},
"negative_invariants": {
"_description": "Alternative A2A channels must be OFF so a passing scenario is only passing via ai-memory shared memory. Any true here = thesis-preserving.",
"a2a_protocol_off": true,
"sub_agent_or_sessions_spawn_off": true,
"alternative_channels_off": true,
"tool_allowlist_is_memory_only": true,
"a2a_gate_profile_locked": true
},
"functional_probes": {
"xai_grok_chat_reachable": true,
"xai_grok_sample_reply": "READY",
"substrate_http_canary_f2a": true,
"substrate_http_canary_uuid": "b51256f6-4e81-4672-b0a7-2c03385ac4c1",
"agent_mcp_canary_f2b": false,
"agent_mcp_canary_uuid": "e4eddad6-f829-41cd-afd0-afb8bfa9a88a",
"agent_canary_response_head": "Traceback (most recent call last): File \"/usr/local/bin/hermes\", line 11, in <module> main() File \"/root/.hermes/hermes-agent/hermes_cli/main.py\", line 8870, in main args.func(args) File \"/root/.hermes/hermes-agent/hermes_cli/main.py\", line 1159, in cmd_chat from cli import main as cli_main File \"/root/.hermes/hermes-agent/cli.py\", line 43, in <module> from prompt_toolkit.history import FileHistory ModuleNotFoundError: No module named 'prompt_toolkit' ",
"_f2b_note": "F2b is LLM-dependent and non-blocking. F2a (deterministic HTTP substrate) gates baseline_pass.",
"mesh_connectivity_f4": true,
"mesh_edges_ok": 3,
"mesh_edges_total": 3,
"mesh_edges_detail": "10.11.0.2:9077:OK,10.11.0.4:9077:OK,10.11.0.5:9077:OK",
"_f4_note": "F4 verifies this local nodes N-1 OUTBOUND mesh edges to every peer via both GET health and POST sync_push dry_run. Aggregator ANDs across N nodes to confirm full N*(N-1) bidirectional reachability. Gates baseline_pass.",
"ai_memory_mcp_stdio_f5": true,
"ai_memory_mcp_stdio_init_ok": true,
"ai_memory_mcp_stdio_tools_ok": true,
"ai_memory_mcp_stdio_tools_found": "memory_agent_list,memory_agent_register,memory_archive_list,memory_archive_purge,memory_archive_restore,memory_archive_stats,memory_auto_tag,memory_capabilities,memory_consolidate,memory_delete,memory_detect_contradiction,memory_expand_query,memory_forget,memory_gc,memory_get,memory_get_links,memory_inbox,memory_link,memory_list,memory_list_subscriptions,memory_namespace_clear_standard,memory_namespace_get_standard,memory_namespace_set_standard,memory_notify,memory_pending_approve,memory_pending_list,memory_pending_reject,memory_promote,memory_recall,memory_search,memory_session_start,memory_stats,memory_store,memory_subscribe,memory_unsubscribe,memory_update",
"_f5_note": "F5 spawns the ai-memory stdio MCP subprocess using the framework-configured invocation and verifies initialize + tools/list return memory_store, memory_recall, memory_list. Deterministic (no LLM). Gates baseline_pass.",
"tls_mode": "off",
"tls_handshake_f6": true,
"tls_handshake_f6_reason": "",
"mtls_enforcement_f7": true,
"mtls_enforcement_f7_reason": "",
"_f6_f7_note": "F6 verifies the TLS 1.3 handshake against the local serve + CA chain. F7 verifies mTLS enforcement — anonymous client rejected, whitelisted client accepted. Both gate baseline_pass when tls_mode != off / mtls respectively.",
"embedder_loaded_f8": true,
"embedder_loaded_f8_reason": "",
"_f8_note": "F8 verifies /api/v1/capabilities reports features.embedder_loaded=true — i.e. the MiniLM embedder initialised at serve startup. Gates baseline_pass unconditionally. Without this, scenario-18 silently black-holes (semantic recall returns 0 rows).",
"agent_mcp_ai_memory_canary": true,
"canary_uuid": "b51256f6-4e81-4672-b0a7-2c03385ac4c1",
"canary_namespace": "_baseline_canary_f2a"
},
"baseline_pass": true
},
{
"spec_version": "1.4.0",
"agent_type": "hermes",
"agent_id": "ai:bob",
"node_index": "2",
"framework_version": "Hermes Agent v0.10.0 (2026.4.16)",
"ai_memory_version": "v0.6.2",
"peer_urls": "http://10.11.0.3:9077,http://10.11.0.4:9077,http://10.11.0.5:9077",
"config_file_sha256": "21635cf6364057fd2a004d28aac89abf8438671d85f9fd2ed1e654d812d23ff1",
"config_attestation": {
"framework_is_authentic": true,
"mcp_server_ai_memory_registered": true,
"llm_backend_is_xai_grok": true,
"llm_is_default_provider": true,
"mcp_command_is_ai_memory": true,
"agent_id_stamped": true,
"federation_live": true,
"ufw_disabled": true,
"iptables_flushed": true,
"dead_man_switch_scheduled": true
},
"negative_invariants": {
"_description": "Alternative A2A channels must be OFF so a passing scenario is only passing via ai-memory shared memory. Any true here = thesis-preserving.",
"a2a_protocol_off": true,
"sub_agent_or_sessions_spawn_off": true,
"alternative_channels_off": true,
"tool_allowlist_is_memory_only": true,
"a2a_gate_profile_locked": true
},
"functional_probes": {
"xai_grok_chat_reachable": true,
"xai_grok_sample_reply": "READY",
"substrate_http_canary_f2a": true,
"substrate_http_canary_uuid": "cf30aaca-e94c-4ad6-8b73-195712065ccd",
"agent_mcp_canary_f2b": false,
"agent_mcp_canary_uuid": "d38e0cac-41d5-4512-ada0-025857b2d83b",
"agent_canary_response_head": "Traceback (most recent call last): File \"/usr/local/bin/hermes\", line 11, in <module> main() File \"/root/.hermes/hermes-agent/hermes_cli/main.py\", line 8870, in main args.func(args) File \"/root/.hermes/hermes-agent/hermes_cli/main.py\", line 1159, in cmd_chat from cli import main as cli_main File \"/root/.hermes/hermes-agent/cli.py\", line 43, in <module> from prompt_toolkit.history import FileHistory ModuleNotFoundError: No module named 'prompt_toolkit' ",
"_f2b_note": "F2b is LLM-dependent and non-blocking. F2a (deterministic HTTP substrate) gates baseline_pass.",
"mesh_connectivity_f4": true,
"mesh_edges_ok": 3,
"mesh_edges_total": 3,
"mesh_edges_detail": "10.11.0.3:9077:OK,10.11.0.4:9077:OK,10.11.0.5:9077:OK",
"_f4_note": "F4 verifies this local nodes N-1 OUTBOUND mesh edges to every peer via both GET health and POST sync_push dry_run. Aggregator ANDs across N nodes to confirm full N*(N-1) bidirectional reachability. Gates baseline_pass.",
"ai_memory_mcp_stdio_f5": true,
"ai_memory_mcp_stdio_init_ok": true,
"ai_memory_mcp_stdio_tools_ok": true,
"ai_memory_mcp_stdio_tools_found": "memory_agent_list,memory_agent_register,memory_archive_list,memory_archive_purge,memory_archive_restore,memory_archive_stats,memory_auto_tag,memory_capabilities,memory_consolidate,memory_delete,memory_detect_contradiction,memory_expand_query,memory_forget,memory_gc,memory_get,memory_get_links,memory_inbox,memory_link,memory_list,memory_list_subscriptions,memory_namespace_clear_standard,memory_namespace_get_standard,memory_namespace_set_standard,memory_notify,memory_pending_approve,memory_pending_list,memory_pending_reject,memory_promote,memory_recall,memory_search,memory_session_start,memory_stats,memory_store,memory_subscribe,memory_unsubscribe,memory_update",
"_f5_note": "F5 spawns the ai-memory stdio MCP subprocess using the framework-configured invocation and verifies initialize + tools/list return memory_store, memory_recall, memory_list. Deterministic (no LLM). Gates baseline_pass.",
"tls_mode": "off",
"tls_handshake_f6": true,
"tls_handshake_f6_reason": "",
"mtls_enforcement_f7": true,
"mtls_enforcement_f7_reason": "",
"_f6_f7_note": "F6 verifies the TLS 1.3 handshake against the local serve + CA chain. F7 verifies mTLS enforcement — anonymous client rejected, whitelisted client accepted. Both gate baseline_pass when tls_mode != off / mtls respectively.",
"embedder_loaded_f8": true,
"embedder_loaded_f8_reason": "",
"_f8_note": "F8 verifies /api/v1/capabilities reports features.embedder_loaded=true — i.e. the MiniLM embedder initialised at serve startup. Gates baseline_pass unconditionally. Without this, scenario-18 silently black-holes (semantic recall returns 0 rows).",
"agent_mcp_ai_memory_canary": true,
"canary_uuid": "cf30aaca-e94c-4ad6-8b73-195712065ccd",
"canary_namespace": "_baseline_canary_f2a"
},
"baseline_pass": true
},
{
"spec_version": "1.4.0",
"agent_type": "hermes",
"agent_id": "ai:charlie",
"node_index": "3",
"framework_version": "Hermes Agent v0.10.0 (2026.4.16)",
"ai_memory_version": "v0.6.2",
"peer_urls": "http://10.11.0.3:9077,http://10.11.0.2:9077,http://10.11.0.5:9077",
"config_file_sha256": "ce52d772ef5a00968db29fb80eea7a14206b0a258a00ff2165db725405474618",
"config_attestation": {
"framework_is_authentic": true,
"mcp_server_ai_memory_registered": true,
"llm_backend_is_xai_grok": true,
"llm_is_default_provider": true,
"mcp_command_is_ai_memory": true,
"agent_id_stamped": true,
"federation_live": true,
"ufw_disabled": true,
"iptables_flushed": true,
"dead_man_switch_scheduled": true
},
"negative_invariants": {
"_description": "Alternative A2A channels must be OFF so a passing scenario is only passing via ai-memory shared memory. Any true here = thesis-preserving.",
"a2a_protocol_off": true,
"sub_agent_or_sessions_spawn_off": true,
"alternative_channels_off": true,
"tool_allowlist_is_memory_only": true,
"a2a_gate_profile_locked": true
},
"functional_probes": {
"xai_grok_chat_reachable": true,
"xai_grok_sample_reply": "READY",
"substrate_http_canary_f2a": true,
"substrate_http_canary_uuid": "2a4635b7-f19b-494d-a097-58e5cf713887",
"agent_mcp_canary_f2b": false,
"agent_mcp_canary_uuid": "8602485a-2d26-4a56-bf7c-87e1fa45d1fe",
"agent_canary_response_head": "Traceback (most recent call last): File \"/usr/local/bin/hermes\", line 11, in <module> main() File \"/root/.hermes/hermes-agent/hermes_cli/main.py\", line 8870, in main args.func(args) File \"/root/.hermes/hermes-agent/hermes_cli/main.py\", line 1159, in cmd_chat from cli import main as cli_main File \"/root/.hermes/hermes-agent/cli.py\", line 43, in <module> from prompt_toolkit.history import FileHistory ModuleNotFoundError: No module named 'prompt_toolkit' ",
"_f2b_note": "F2b is LLM-dependent and non-blocking. F2a (deterministic HTTP substrate) gates baseline_pass.",
"mesh_connectivity_f4": true,
"mesh_edges_ok": 3,
"mesh_edges_total": 3,
"mesh_edges_detail": "10.11.0.3:9077:OK,10.11.0.2:9077:OK,10.11.0.5:9077:OK",
"_f4_note": "F4 verifies this local nodes N-1 OUTBOUND mesh edges to every peer via both GET health and POST sync_push dry_run. Aggregator ANDs across N nodes to confirm full N*(N-1) bidirectional reachability. Gates baseline_pass.",
"ai_memory_mcp_stdio_f5": true,
"ai_memory_mcp_stdio_init_ok": true,
"ai_memory_mcp_stdio_tools_ok": true,
"ai_memory_mcp_stdio_tools_found": "memory_agent_list,memory_agent_register,memory_archive_list,memory_archive_purge,memory_archive_restore,memory_archive_stats,memory_auto_tag,memory_capabilities,memory_consolidate,memory_delete,memory_detect_contradiction,memory_expand_query,memory_forget,memory_gc,memory_get,memory_get_links,memory_inbox,memory_link,memory_list,memory_list_subscriptions,memory_namespace_clear_standard,memory_namespace_get_standard,memory_namespace_set_standard,memory_notify,memory_pending_approve,memory_pending_list,memory_pending_reject,memory_promote,memory_recall,memory_search,memory_session_start,memory_stats,memory_store,memory_subscribe,memory_unsubscribe,memory_update",
"_f5_note": "F5 spawns the ai-memory stdio MCP subprocess using the framework-configured invocation and verifies initialize + tools/list return memory_store, memory_recall, memory_list. Deterministic (no LLM). Gates baseline_pass.",
"tls_mode": "off",
"tls_handshake_f6": true,
"tls_handshake_f6_reason": "",
"mtls_enforcement_f7": true,
"mtls_enforcement_f7_reason": "",
"_f6_f7_note": "F6 verifies the TLS 1.3 handshake against the local serve + CA chain. F7 verifies mTLS enforcement — anonymous client rejected, whitelisted client accepted. Both gate baseline_pass when tls_mode != off / mtls respectively.",
"embedder_loaded_f8": true,
"embedder_loaded_f8_reason": "",
"_f8_note": "F8 verifies /api/v1/capabilities reports features.embedder_loaded=true — i.e. the MiniLM embedder initialised at serve startup. Gates baseline_pass unconditionally. Without this, scenario-18 silently black-holes (semantic recall returns 0 rows).",
"agent_mcp_ai_memory_canary": true,
"canary_uuid": "2a4635b7-f19b-494d-a097-58e5cf713887",
"canary_namespace": "_baseline_canary_f2a"
},
"baseline_pass": true
}
]
}
Workflow-level probe answering "can agents communicate through ai-memory?". Writer ai:alice posted canary UUID c529b158-1394-4992-8ca7-60149ca44c5e to namespace _baseline_peer_canary via node-1's local ai-memory serve HTTP. After W=2 fanout settle, probe confirmed the canary on each of the 3 peer nodes via their local GET /api/v1/memories.
{
"probe": "F3",
"name": "peer-a2a-via-shared-memory",
"description": "Writer agent posts a canary via local ai-memory HTTP on node-1; verifies the row propagates to the 3 peer nodes (W=2/N=4 quorum) before scenarios run.",
"canary_uuid": "c529b158-1394-4992-8ca7-60149ca44c5e",
"canary_namespace": "_baseline_peer_canary",
"writer_agent": "ai:alice",
"pass": true
}
Run focus
What this campaign tested: Exercised 34 scenarios covering basic recall, handoffs, deletions, links, registrations, versioning, partitions, promotions, queries, bulk ops, and advanced features like consolidation, contradictions, archiving, notifications, and exports over HTTP transport in a 4-node federation without TLS.
What it demonstrated: Demonstrated reliable memory sharing and replication across agents in most tests, but revealed a failure in semantic query recall for one agent's memory and a skipped scenario due to unparseable report.
AI NHI analysis · Claude Opus 4.7
PARTIAL — 33/34 scenarios passed, S18 failed, S23 skipped.
The AI agents were able to share and remember information with each other reliably in almost all tests. However, in one case, a search for similar ideas missed something that should have been found. One other test was skipped because its results couldn't be read properly.
Risk posture remains low with high reliability in core memory sharing, supporting production readiness for most features; customer claims on semantic search viability need caution due to intermittent misses. This patch maintains stability versus prior runs, with the semantic flake persisting as a known issue. No regressions noted, but skipped scenario requires fix for full coverage.
Failure in S18 where semantic query on 'morning outdoor exercise routine' did not surface Bob's memory (seen_by_charlie=0), likely due to embedding similarity threshold or index flake; impacts hybrid_recall primitive. S23 skipped from unparseable JSON, probable harness bug in report generation. Other primitives like keyword_search (S28), bulk insert (S40), and delta queries (S39) passed cleanly across nodes.
Investigate and fix S18 semantic recall flake, and resolve S23 report parsing issue for complete coverage.
Every scenario that produced a JSON report in this campaign, in testbook order. Click a row's scenario id to jump to its full report below. See the Every test performed page for the authoritative catalog.
| ID | Title | Result | Reason |
|---|---|---|---|
| S1 | Per-agent write + read (MCP stdio) | PASS | |
| S1b | Per-agent write + read (HTTP) | PASS | |
| S2 | Shared-context handoff | PASS | |
| S4 | Federation-aware concurrent writes | PASS | |
| S5 | Consolidation + curation | PASS | |
| S6 | Contradiction detection | PASS | |
| S9 | Mutation round-trip | PASS | |
| S10 | Deletion propagation | PASS | |
| S11 | Link integrity | PASS | |
| S12 | Agent registration | PASS | |
| S13 | Concurrent write contention | PASS | |
| S14 | Partition tolerance | PASS | |
| S15 | Read-your-writes | PASS | |
| S16 | Tier promotion | PASS | |
| S17 | Stats consistency | PASS | |
| S18 | Semantic query expansion | ? | semantic query did not surface bob's memory |
| S22 | Identity spoofing resistance | PASS | |
| S23 | Malicious content fuzz | ? | |
| S24 | Byzantine peer | PASS | |
| S25 | Clock skew tolerance | PASS | |
| S28 | memory_search keyword | PASS | |
| S29 | memory_archive lifecycle | PASS | |
| S30 | memory_capabilities handshake | PASS | |
| S31 | memory_gc quiescence | PASS | |
| S32 | memory_inbox + notify | PASS | |
| S33 | memory_subscribe pub/sub | PASS | |
| S34 | memory_pending governance | PASS | |
| S35 | memory_namespace standards | PASS | |
| S36 | memory_session_start | PASS | |
| S37 | memory_get_links bidirectional | PASS | |
| S38 | /export + /import | PASS | |
| S39 | /sync/since delta | PASS | |
| S40 | /memories/bulk | PASS | |
| S41 | /metrics Prometheus | PASS | |
| S42 | /namespaces enumeration | PASS |
{
"agent_group": "hermes",
"expected_per_reader": 20,
"pass": true,
"per_agent": {
"ai:alice": {
"recall": 20
},
"ai:bob": {
"recall": 20
},
"ai:charlie": {
"recall": 20
}
},
"per_namespace_node4": {
"scenario1-ai:alice": {
"count": 10,
"wrong_agent_id": 0
},
"scenario1-ai:bob": {
"count": 10,
"wrong_agent_id": 0
},
"scenario1-ai:charlie": {
"count": 10,
"wrong_agent_id": 0
}
},
"reasons": [],
"scenario": "1",
"skipped": false,
"tls_mode": "off"
}
phase A: each agent writes 10 memories via MCP ai:alice on 104.236.92.237 ai:bob on 104.236.54.83 ai:charlie on 104.236.96.109 settle 15s for W=2/N=4 convergence phase B: each agent counts rows in the OTHER two namespaces ai:alice recalled 20 rows from the other two namespaces ai:bob recalled 20 rows from the other two namespaces ai:charlie recalled 20 rows from the other two namespaces phase C: cross-cluster identity check on node-4 ns=scenario1-ai:alice count=10 wrong_agent_id=0 ns=scenario1-ai:bob count=10 wrong_agent_id=0 ns=scenario1-ai:charlie count=10 wrong_agent_id=0
{
"agent_group": "hermes",
"expected_per_reader": 20,
"pass": true,
"path": "serve-http",
"per_agent": {
"ai:alice": {
"recall": 20
},
"ai:bob": {
"recall": 20
},
"ai:charlie": {
"recall": 20
}
},
"per_namespace_node4": {
"scenario1b-ai:alice": {
"count": 10,
"wrong_agent_id": 0
},
"scenario1b-ai:bob": {
"count": 10,
"wrong_agent_id": 0
},
"scenario1b-ai:charlie": {
"count": 10,
"wrong_agent_id": 0
}
},
"reasons": [],
"scenario": "1b",
"skipped": false,
"tls_mode": "off"
}
phase A: each agent POSTs 10 memories to local serve ai:alice on 104.236.92.237 ai:bob on 104.236.54.83 ai:charlie on 104.236.96.109 settle 15s for W=2/N=4 convergence phase B: count rows in other two namespaces via local serve HTTP ai:alice sees 20 rows from the other two namespaces ai:bob sees 20 rows from the other two namespaces ai:charlie sees 20 rows from the other two namespaces phase C: cross-cluster identity check on node-4 ns=scenario1b-ai:alice count=10 wrong_agent_id=0 ns=scenario1b-ai:bob count=10 wrong_agent_id=0 ns=scenario1b-ai:charlie count=10 wrong_agent_id=0
{
"ack_uuid": "a-58f570f27cfd4e84843a204420b89e13",
"agent_group": "hermes",
"handoff_uuid": "h-21364f8365ad4d47bb0f26ed033dd971",
"pass": true,
"path": "serve-http",
"per_agent": {
"ai:alice": {
"sees_ack": 1
},
"ai:bob": {
"sees_handoff": 1
}
},
"reasons": [],
"scenario": "2",
"skipped": false,
"tls_mode": "off"
}
phase A: ai:alice writes handoff to ai:bob (uuid=h-21364f8365ad4d47bb0f26ed033dd971) settle 8s for quorum fanout phase B: ai:bob reads handoff on node-2 ai:bob sees 1 handoff memories from ai:alice phase C: ai:bob writes acknowledgement (uuid=a-58f570f27cfd4e84843a204420b89e13) settle 8s for reverse-direction fanout phase D: ai:alice reads ack on node-1 ai:alice sees 1 ack memories from ai:bob
{
"agent_group": "hermes",
"expected_per_agent": 30,
"pass": true,
"per_agent": {
"ai:alice": {
"count": 30,
"wrong_agent_id": 0
},
"ai:bob": {
"count": 30,
"wrong_agent_id": 0
},
"ai:charlie": {
"count": 30,
"wrong_agent_id": 0
}
},
"reasons": [],
"scenario": "4",
"skipped": false,
"tls_mode": "off"
}
phase A: launching concurrent 30-row bursts from 3 agents ai:alice burst ok=30/30 ai:bob burst ok=30/30 ai:charlie burst ok=30/30 settle 20s for W=2 fanout convergence phase B: querying node-4 aggregator for per-agent counts ai:alice: count=30 (expected 30) wrong_agent_id=0 ai:bob: count=30 (expected 30) wrong_agent_id=0 ai:charlie: count=30 (expected 30) wrong_agent_id=0
{
"agent_group": "hermes",
"consolidate_http_code": 201,
"consolidated_from_agents": [
"ai:charlie",
"ai:bob",
"ai:alice"
],
"consolidated_id": "fb89731a-64ee-477d-924f-c80bc02e0583",
"pass": true,
"reasons": [],
"scenario": "5",
"skipped": false,
"tls_mode": "off"
}
phase A: each agent writes 3 related memories ai:alice on 104.236.92.237 ai:bob on 104.236.54.83 ai:charlie on 104.236.96.109 settle 8s for quorum fanout phase B: collect source ids on node-1, then trigger consolidate source ids (count=9): ['a19a675f-8f2f-4fcd-afe9-de74c778c629', 'f983fda7-8302-45e6-a8a4-535fa6b4aebb', '68bdb168-0391-4c1e-b171-9460b3eb2a4f', 'ec2c63fd-7afb-413c-a5c1-e934fa91ded7', 'aaddc604-ab77-4f6d-a1e4-3ac7b8ec958b']... consolidate HTTP 201, consolidated_id=fb89731a-64ee-477d-924f-c80bc02e0583 settle 10s for consolidation fanout phase C: verifying consolidated_from_agents on node-4 consolidated_from_agents=['ai:charlie', 'ai:bob', 'ai:alice']
{
"agent_group": "hermes",
"alice_id": "9d221694-4777-44a7-954c-6705f3c179ed",
"bob_id": "d927f53a-2193-43a5-b298-95638e3e1e1a",
"charlie_sees_both_memories": true,
"charlie_sees_contradicts_link": true,
"detect_http_code": 200,
"pass": true,
"reasons": [],
"scenario": "6",
"skipped": false,
"tls_mode": "off",
"topic": "sky-color-8c645a7f"
}
alice writes claim: "sky-color-8c645a7f is blue" on node-1 bob writes contradicting claim: "sky-color-8c645a7f is red" on node-2 alice.id=9d221694-4777-44a7-954c-6705f3c179ed bob.id=d927f53a-2193-43a5-b298-95638e3e1e1a settle 10s for quorum fanout + contradiction indexing charlie queries /api/v1/contradictions on node-3 HTTP 200 sees both memories: True; sees contradicts link: True
{
"agent_group": "hermes",
"charlie_view": {
"agent_id": "ai:alice",
"content": "v2-25f7e57a7908468fb89f2fa094018f19"
},
"m1_id": "b02c2075-0b3d-4074-840c-38d0bb8c1df9",
"pass": true,
"put_http_code": 200,
"reasons": [],
"scenario": "9",
"skipped": false,
"tls_mode": "off",
"v1_uuid": "v1-7864325464ae400ba4b3305697359fdf",
"v2_uuid": "v2-25f7e57a7908468fb89f2fa094018f19"
}
alice writes M1 content=v1-7864325464ae400ba4b3305697359fdf on node-1 M1 id=b02c2075-0b3d-4074-840c-38d0bb8c1df9 settle 5s for initial replication bob updates M1 content=v2-25f7e57a7908468fb89f2fa094018f19 on node-2 via PUT PUT returned HTTP 200 settle 8s for update fanout charlie reads M1 on node-3 and checks content + provenance charlie sees content="v2-25f7e57a7908468fb89f2fa094018f19" agent_id="ai:alice"
{
"agent_group": "hermes",
"delete_http_code": 200,
"m1_id": "924b41e3-fb5a-473c-a93c-5b5b0b3ce592",
"pass": true,
"post_delete_hits": {
"node-2": 0,
"node-3": 0,
"node-4": 0
},
"post_delete_still_visible_peers": 0,
"pre_delete_visible_peers": 3,
"reasons": [],
"scenario": "10",
"skipped": false,
"tls_mode": "off",
"uuid": "d-6d4be220ce6c4496a7948e46233e46aa"
}
alice writes M1 content=d-6d4be220ce6c4496a7948e46233e46aa on node-1 created memory id=924b41e3-fb5a-473c-a93c-5b5b0b3ce592 settle 8s for pre-delete fanout pre-delete: verifying M1 is visible on all peers pre-delete node-2 sees 1 pre-delete node-3 sees 1 pre-delete node-4 sees 1 alice deletes M1 on node-1 DELETE returned HTTP 200 settle 15s for tombstone propagation post-delete: verifying M1 is GONE from all peers post-delete node-2 sees 0 (expected 0) post-delete node-3 sees 0 (expected 0) post-delete node-4 sees 0 (expected 0)
{
"agent_group": "hermes",
"charlie_sees_link": 1,
"link_http_code": 201,
"m1_id": "4326350a-2c34-4509-9a47-2f9cab41a146",
"m2_id": "d5158bd6-4ad1-49fc-ab44-093d235e821b",
"pass": true,
"reasons": [],
"relation": "related_to",
"scenario": "11",
"skipped": false,
"tls_mode": "off"
}
alice writes M1 on node-1 bob writes M2 on node-2 M1=4326350a-2c34-4509-9a47-2f9cab41a146 M2=d5158bd6-4ad1-49fc-ab44-093d235e821b settle 5s for pre-link replication alice links M1 -> M2 with relation=related_to link POST returned HTTP 201 settle 8s for link fanout charlie queries links of M1 on node-3 charlie sees M1->M2 link: 1 (expected >=1)
{
"agent_group": "hermes",
"pass": true,
"peers_see": {
"node_2": 1,
"node_3": 1,
"node_4": 1
},
"reasons": [],
"register_http_code": 201,
"registered_agent": "ai:dave-probe-1eefc6ba",
"scenario": "12",
"skipped": false,
"tls_mode": "off"
}
alice registers new agent ai:dave-probe-1eefc6ba on node-1 POST /api/v1/agents returned HTTP 201 settle 10s for agent-list fanout node-2 sees ai:dave-probe-1eefc6ba: 1 (expected >=1) node-3 sees ai:dave-probe-1eefc6ba: 1 (expected >=1) node-4 sees ai:dave-probe-1eefc6ba: 1 (expected >=1)
{
"agent_group": "hermes",
"m1_id": "c0ae4256-034b-48a8-a97f-61b4b55947fd",
"pass": true,
"peer_view": {
"node_1": "vb-ffa01bf129674c7f8b90750e10368147",
"node_2": "vb-ffa01bf129674c7f8b90750e10368147",
"node_3": "vb-ffa01bf129674c7f8b90750e10368147",
"node_4": "vb-ffa01bf129674c7f8b90750e10368147"
},
"reasons": [],
"scenario": "13",
"skipped": false,
"submitted": {
"v0": "v0-173527d3ed994018824681d53987c7ea",
"vA_alice": "va-a03f1cc3197a47a3a9f55d221e470fe2",
"vB_bob": "vb-ffa01bf129674c7f8b90750e10368147"
},
"tls_mode": "off"
}
alice writes M1 content=v0-173527d3ed994018824681d53987c7ea on node-1
M1 id=c0ae4256-034b-48a8-a97f-61b4b55947fd
settle 5s for initial replication
alice + bob issue concurrent PUTs (vA=va-a03f1cc3197a47a3a9f55d221e470fe2 from alice, vB=vb-ffa01bf129674c7f8b90750e10368147 from bob)
concurrent PUT results: [(0, {'body': {'access_count': 0, 'confidence': 1.0, 'content': 'va-a03f1cc3197a47a3a9f55d221e470fe2', 'created_at': '2026-04-23T20:37:03.866706087+00:00', 'expires_at': '2026-04-30T20:37:03.866706087+00:00', 'id': 'c0ae4256-034b-48a8-a97f-61b4b55947fd', 'metadata': {'agent_id': 'ai:alice', 'scenario': '13'}, 'namespace': 'scenario13-contention', 'priority': 5, 'source': 'api', 'tags': [], 'tier': 'mid', 'title': 'm1', 'updated_at': '2026-04-23T20:37:10.074855063+00:00'}, 'http_code': 200}), (0, {'body': {'access_count': 0, 'confidence': 1.0, 'content': 'vb-ffa01bf129674c7f8b90750e10368147', 'created_at': '2026-04-23T20:37:03.866706087+00:00', 'expires_at': '2026-04-30T20:37:03.866706087+00:00', 'id': 'c0ae4256-034b-48a8-a97f-61b4b55947fd', 'metadata': {'agent_id': 'ai:alice', 'scenario': '13'}, 'namespace': 'scenario13-contention', 'priority': 5, 'source': 'api', 'tags': [], 'tier': 'mid', 'title': 'm1', 'updated_at': '2026-04-23T20:37:10.975496124+00:00'}, 'http_code': 200})]
settle 10s for quorum convergence
node-1 sees content=vb-ffa01bf129674c7f8b90750e10368147
node-2 sees content=vb-ffa01bf129674c7f8b90750e10368147
node-3 sees content=vb-ffa01bf129674c7f8b90750e10368147
node-4 sees content=vb-ffa01bf129674c7f8b90750e10368147
{
"agent_group": "hermes",
"expected_post_recovery": 20,
"node3_saw": 20,
"partition_target": "node-3",
"pass": true,
"reasons": [],
"scenario": "14",
"skipped": false,
"tls_mode": "off"
}
suspending ai-memory on node-3 (SIGSTOP) !! ssh timeout (15s): root@104.236.96.109 pgrep -f 'ai-memory serve' | xargs -r kill -STOP settle 2s for process-suspend observe writing 10 memories each from alice + bob during node-3 outage resuming ai-memory on node-3 (SIGCONT) settle 20s for post-partition catchup checking node-3 caught up node-3 sees 20 memories in scenario14-partition (expected 20)
{
"agent_group": "hermes",
"pass": true,
"reasons": [],
"scenario": "15",
"skipped": false,
"tls_mode": "off",
"uuid": "ryw-51645efa28104d8fbf9c50d46dd0dfdd",
"writer_sees_own_write": 1
}
alice writes + immediately reads M1 on node-1 (uuid=ryw-51645efa28104d8fbf9c50d46dd0dfdd) alice sees 1 (expected 1) immediately after write
{
"agent_group": "hermes",
"bob_sees_tier": "long",
"m1_id": "8be2da31-b358-440f-80f4-74cfced7acda",
"pass": true,
"promote_http_code": 200,
"reasons": [],
"scenario": "16",
"skipped": false,
"tls_mode": "off"
}
alice writes M1 tier=short on node-1 M1 id=8be2da31-b358-440f-80f4-74cfced7acda settle 5s for pre-promote replication alice promotes M1 to tier=long promote returned HTTP 200 settle 8s for promotion fanout bob sees tier=long (expected long)
{
"agent_group": "hermes",
"expected_count": 15,
"pass": true,
"per_peer": {
"node_1": 15,
"node_2": 15,
"node_3": 15,
"node_4": 15
},
"reasons": [],
"scenario": "17",
"skipped": false,
"tls_mode": "off"
}
phase A: each of 3 agents writes 5 memories to scenario17-stats ai:alice on 104.236.92.237 ai:bob on 104.236.54.83 ai:charlie on 104.236.96.109 settle 15s for W=2 fanout phase B: querying count on every peer node-1 count=15 (expected 15) node-2 count=15 (expected 15) node-3 count=15 (expected 15) node-4 count=15 (expected 15)
Reasons: semantic query did not surface bob's memory
{
"agent_group": "hermes",
"pass": false,
"query": "morning outdoor exercise routine",
"reason": "semantic query did not surface bob's memory",
"reasons": [
"semantic query did not surface bob's memory"
],
"scenario": "18",
"skipped": false,
"tls_mode": "off",
"writers": [
{
"agent": "ai:alice",
"marker": "alice-sunrise-1857dd4f",
"seen_by_charlie": 1
},
{
"agent": "ai:bob",
"marker": "bob-daybreak-00784da2",
"seen_by_charlie": 0
}
]
}
alice writes A on node-1 bob writes B on node-2 settle 15s for fanout + index rebuild charlie queries on node-3 with semantically-related prompt charlie sees alice's memory: 1 (expected >=1) charlie sees bob's memory: 0 (expected >=1)
{
"agent_group": "hermes",
"pass": true,
"reasons": [],
"scenario": "22",
"skipped": false,
"tests": {
"body_vs_header_conflict": {
"acceptable": [
"ai:body-wins",
"ai:attacker"
],
"stored_agent_id": "ai:attacker"
},
"header_only": {
"expected": "ai:alice",
"stored_agent_id": "ai:alice"
}
},
"tls_mode": "off"
}
test 1: header-only X-Agent-Id=ai:alice settle 2s for read-settle stored metadata.agent_id for header-only write: ai:alice (expected ai:alice) test 2: body.metadata.agent_id=ai:body-wins vs X-Agent-Id=ai:attacker settle 2s for read-settle stored metadata.agent_id for body+header conflict: ai:attacker
payload sql: 61 bytes
payload html: 66 bytes
payload oversize: 1048576 bytes
Traceback (most recent call last):
File "/home/runner/work/ai-memory-ai2ai-gate/ai-memory-ai2ai-gate/scripts/scenarios/23_malicious_content_fuzz.py", line 106, in <module>
main()
File "/home/runner/work/ai-memory-ai2ai-gate/ai-memory-ai2ai-gate/scripts/scenarios/23_malicious_content_fuzz.py", line 49, in main
rc, write_doc = h.write_memory(
^^^^^^^^^^^^^^^
File "/home/runner/work/ai-memory-ai2ai-gate/ai-memory-ai2ai-gate/scripts/a2a_harness.py", line 218, in write_memory
return self.http_on(node_ip, "POST", "/api/v1/memories",
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/runner/work/ai-memory-ai2ai-gate/ai-memory-ai2ai-gate/scripts/a2a_harness.py", line 173, in http_on
result = self.ssh_exec(node_ip, remote_cmd, timeout=timeout)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/runner/work/ai-memory-ai2ai-gate/ai-memory-ai2ai-gate/scripts/a2a_harness.py", line 119, in ssh_exec
return self._run(cmd, timeout=timeout, stdin=stdin)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/runner/work/ai-memory-ai2ai-gate/ai-memory-ai2ai-gate/scripts/a2a_harness.py", line 103, in _run
return subprocess.run(
^^^^^^^^^^^^^^^
File "/usr/lib/python3.12/subprocess.py", line 548, in run
with Popen(*popenargs, **kwargs) as process:
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/lib/python3.12/subprocess.py", line 1026, in __init__
self._execute_child(args, executable, preexec_fn, close_fds,
File "/usr/lib/python3.12/subprocess.py", line 1955, in _execute_child
raise child_exception_type(errno_num, err_msg, err_filename)
OSError: [Errno 7] Argument list too long: 'ssh'
{
"agent_group": "hermes",
"byzantine_marker": "bz-ffa067f386554def8f421214efa2b76e",
"pass": true,
"reasons": [],
"scenario": "24",
"skipped": false,
"stored_metadata_agent_id": "REJECTED_BY_SERVER",
"sync_push_http_code": "422",
"tls_mode": "off"
}
node-2 sends sync_push to node-3 claiming sender_agent_id=ai:alice sync_push returned HTTP 422 settle 5s for server-side sync apply node-3 stored metadata.agent_id=ABSENT (declared: ai:alice) sync_push rejected HTTP 422 — stricter-than-spec, acceptable
{
"agent_group": "hermes",
"clock_offset_seconds": 300,
"marker": "ck-ae565070642a4c54bd53f4abc03e337a",
"pass": true,
"reasons": [],
"scenario": "25",
"seen_on": {
"node_1": 1,
"node_3": 1
},
"skipped": false,
"target_node": "node-3",
"tls_mode": "off"
}
shifting node-3 clock +300s (NTP disabled for the duration) node-3 now reports: Thu Apr 23 20:45:03 UTC 2026 alice writes on node-1 (normal clock); waiting for quorum fanout to skewed node-3 settle 15s for skewed-peer convergence node-3 (+300s clock) sees marker: 1 (expected >=1) node-1 sees marker: 1 (expected >=1) reverting node-3 clock
{
"agent_group": "hermes",
"pass": true,
"peer_hits": {
"node_2": 1,
"node_3": 1
},
"reasons": [],
"scenario": "28",
"skipped": false,
"tls_mode": "off",
"token": "kwsearch260d6bdbe7"
}
alice writes a row containing unique token=kwsearch260d6bdbe7 settle 8s for search index populate + fanout bob + charlie call /api/v1/search with the exact token node-2 keyword search returned 1 hits node-3 keyword search returned 1 hits
{
"agent_group": "hermes",
"archive_http_code": 200,
"bob_sees_archived": true,
"m1_id": "f4647989-b842-42bf-bb3f-504f97aa122e",
"node4_active_rows": 1,
"pass": true,
"reasons": [],
"restore_http_code": 200,
"scenario": "29",
"skipped": false,
"stats_shape_ok": true,
"tls_mode": "off"
}
alice writes M1 on node-1
M1 id=f4647989-b842-42bf-bb3f-504f97aa122e
settle 5s for pre-archive replication
alice archives M1 via POST /api/v1/archive (ai-memory-mcp PR #361)
archive (POST) returned HTTP 200
settle 5s for archive propagation
bob queries /api/v1/archive on node-2
bob sees M1 in archive: True
charlie restores M1 via /api/v1/archive/{id}/restore on node-3
restore returned HTTP 200
settle 5s for restore propagation
node-4 aggregator: M1 must be active again
node-4 active rows matching marker: 1
fetch /api/v1/archive/stats on node-4
{
"agent_group": "hermes",
"pass": true,
"peer_views": {
"node_1": {
"_path": "/api/v1/capabilities",
"features": {
"auto_consolidation": false,
"auto_tagging": false,
"contradiction_analysis": false,
"cross_encoder_reranking": false,
"embedder_loaded": true,
"hybrid_recall": true,
"keyword_search": true,
"memory_reflection": false,
"query_expansion": false,
"semantic_search": true
},
"models": {
"cross_encoder": "none",
"embedding": "sentence-transformers/all-MiniLM-L6-v2",
"embedding_dim": 384,
"llm": "none"
},
"tier": "semantic",
"version": "0.6.2"
},
"node_2": {
"_path": "/api/v1/capabilities",
"features": {
"auto_consolidation": false,
"auto_tagging": false,
"contradiction_analysis": false,
"cross_encoder_reranking": false,
"embedder_loaded": true,
"hybrid_recall": true,
"keyword_search": true,
"memory_reflection": false,
"query_expansion": false,
"semantic_search": true
},
"models": {
"cross_encoder": "none",
"embedding": "sentence-transformers/all-MiniLM-L6-v2",
"embedding_dim": 384,
"llm": "none"
},
"tier": "semantic",
"version": "0.6.2"
},
"node_3": {
"_path": "/api/v1/capabilities",
"features": {
"auto_consolidation": false,
"auto_tagging": false,
"contradiction_analysis": false,
"cross_encoder_reranking": false,
"embedder_loaded": true,
"hybrid_recall": true,
"keyword_search": true,
"memory_reflection": false,
"query_expansion": false,
"semantic_search": true
},
"models": {
"cross_encoder": "none",
"embedding": "sentence-transformers/all-MiniLM-L6-v2",
"embedding_dim": 384,
"llm": "none"
},
"tier": "semantic",
"version": "0.6.2"
},
"node_4": {
"_path": "/api/v1/capabilities",
"features": {
"auto_consolidation": false,
"auto_tagging": false,
"contradiction_analysis": false,
"cross_encoder_reranking": false,
"embedder_loaded": true,
"hybrid_recall": true,
"keyword_search": true,
"memory_reflection": false,
"query_expansion": false,
"semantic_search": true
},
"models": {
"cross_encoder": "none",
"embedding": "sentence-transformers/all-MiniLM-L6-v2",
"embedding_dim": 384,
"llm": "none"
},
"tier": "semantic",
"version": "0.6.2"
}
},
"reasons": [],
"scenario": "30",
"skipped": false,
"tls_mode": "off"
}
node-1 capabilities: ['features', 'models', 'tier', 'version', '_path'] node-2 capabilities: ['features', 'models', 'tier', 'version', '_path'] node-3 capabilities: ['features', 'models', 'tier', 'version', '_path'] node-4 capabilities: ['features', 'models', 'tier', 'version', '_path']
{
"agent_group": "hermes",
"expected_live": 2,
"forget_http_code": 400,
"gc_http_code": 200,
"live_markers_per_peer": {
"node_1": 2,
"node_2": 2,
"node_3": 2,
"node_4": 2
},
"pass": true,
"reasons": [],
"scenario": "31",
"skipped": false,
"tls_mode": "off"
}
alice writes 4 memories settle 6s for pre-gc replication alice forgets 2 via /api/v1/forget forget returned HTTP 400 settle 5s for forget propagation bob triggers /api/v1/gc on node-2 gc returned HTTP 200 settle 8s for post-gc settle verify remaining 2 markers are still readable on every peer node-1 sees 2/2 live markers node-2 sees 2/2 live markers node-3 sees 2/2 live markers node-4 sees 2/2 live markers
{
"agent_group": "hermes",
"bob_inbox_count": 1,
"bob_sees_marker": true,
"charlie_inbox_count": 0,
"charlie_sees_marker": false,
"marker": "inb-2985b41d1b7342b7ae44982d198c2669",
"notify_http_code": 201,
"pass": true,
"reasons": [],
"scenario": "32",
"skipped": false,
"tls_mode": "off"
}
alice calls /api/v1/notify → target=ai:bob notify returned HTTP 201 settle 6s for notification fanout bob queries his inbox on node-2 bob inbox has 1 messages; sees marker: True charlie queries his inbox on node-3 (must NOT see it) charlie inbox has 0 messages; sees marker: False
{
"agent_group": "hermes",
"m1_delivered": 1,
"namespace": "scenario33-pubsub-dc140f",
"ns_in_subs_after": false,
"ns_in_subs_before": true,
"pass": true,
"reasons": [],
"scenario": "33",
"skipped": false,
"subscribe_http_code": 201,
"subscriptions_after_count": 0,
"subscriptions_before_count": 1,
"tls_mode": "off",
"unsubscribe_http_code": 200
}
bob subscribes to namespace scenario33-pubsub-dc140f on node-2 subscribe returned HTTP 201 settle 2s for subscription settle bob subscriptions: 1 entries; contains ns: True alice writes M1 into the subscribed namespace settle 6s for write fanout to subscribers bob sees M1 in subscribed namespace: 1 bob unsubscribes from scenario33-pubsub-dc140f unsubscribe returned HTTP 200 settle 2s for unsubscribe settle bob subscriptions after unsubscribe: ns still present = False alice writes M2 post-unsubscribe (may still replicate via federation but subscription list excludes ns) settle 5s for post-unsubscribe settle
{
"agent_group": "hermes",
"approve_http_code": 200,
"charlie_sees": {
"approved": 1,
"rejected": 0
},
"namespace": "scenario34-pending-6ba435",
"pass": true,
"pending_queue_count": 2,
"reasons": [],
"reject_http_code": 200,
"scenario": "34",
"set_standard_http_code": 201,
"skipped": false,
"tls_mode": "off"
}
alice sets namespace standard on scenario34-pending-6ba435: write=approve, approver=ai:bob set-standard returned HTTP 201 settle 2s for standard settle alice writes two memories into the governed namespace (should land in pending) p1=432b7865-9f4c-4c7d-a7bf-4e4b4a840fa4 p2=e5194845-a59e-4fb8-b33a-a232355f4b92 settle 4s for pending queue settle bob lists pending on node-2 pending queue has 2 entries bob approves p1, rejects p2 approve HTTP 200; reject HTTP 200 settle 5s for decision fanout charlie reads the namespace — expects ONLY approved marker charlie sees approved=1 rejected=0
{
"agent_group": "hermes",
"child_ns": "scenario35-parent-a20d37/child",
"clear_http_code": 200,
"get_standard_http_code": 200,
"parent_ns": "scenario35-parent-a20d37",
"pass": true,
"post_clear_has_child_rule": false,
"reasons": [],
"scenario": "35",
"sees_child_rule": true,
"sees_parent_rule": true,
"set_child_http_code": 201,
"set_parent_http_code": 201,
"skipped": false,
"tls_mode": "off"
}
alice writes parent-standard-memory on node-1 alice sets namespace standard on scenario35-parent-a20d37 set-parent returned HTTP 201 alice writes child-standard-memory on node-1 alice sets namespace standard on scenario35-parent-a20d37/child with parent=scenario35-parent-a20d37 set-child returned HTTP 201 settle 4s for standard fanout bob gets standard for scenario35-parent-a20d37/child on node-2 (expects layered parent+child) get-standard returned HTTP 200 parent-rule visible=True; child-rule visible=True alice clears standard on scenario35-parent-a20d37/child clear returned HTTP 200 settle 3s for clear settle
{
"agent_group": "hermes",
"pass": true,
"reasons": [],
"scenario": "36",
"session_id": "90a10de7-e671-4654-ba8a-7b8f6970f0c6",
"session_tagged_rows_on_bob": 2,
"skipped": false,
"start_http_code": 200,
"tls_mode": "off"
}
alice starts a session on node-1 session_start returned HTTP 200, session_id=90a10de7-e671-4654-ba8a-7b8f6970f0c6 alice writes 2 memories tagged with session_id settle 6s for session-tagged fanout bob lists on node-2 filtered by session_id=90a10de7-e671-4654-ba8a-7b8f6970f0c6 bob sees 2 rows tagged session_id=90a10de7-e671-4654-ba8a-7b8f6970f0c6 (expected 2)
{
"agent_group": "hermes",
"forward_has_target": true,
"m1": "eaf5b702-8f06-4bec-8d19-9eb1a8beeb54",
"m2": "c1431f4c-4d69-41a7-ab27-41c95bc0fca0",
"pass": true,
"reasons": [],
"reverse_has_source": true,
"scenario": "37",
"skipped": false,
"tls_mode": "off"
}
alice writes M1 + M2 + links M1→M2 M1=eaf5b702-8f06-4bec-8d19-9eb1a8beeb54 M2=c1431f4c-4d69-41a7-ab27-41c95bc0fca0 settle 6s for link fanout charlie queries /api/v1/links/M1 (forward) charlie queries /api/v1/links/M2 (reverse)
{
"agent_group": "hermes",
"dst_ns": "scenario38-dst-fd1b46",
"expected_rows": 5,
"export_http_code": 200,
"import_http_code": 200,
"markers_preserved": 5,
"pass": true,
"reasons": [],
"rows_exported": 5,
"rows_in_destination": 5,
"scenario": "38",
"skipped": false,
"src_ns": "scenario38-src-fd1b46",
"tls_mode": "off"
}
alice writes 5 rows into scenario38-src-fd1b46 settle 4s for pre-export replication alice exports on node-1 (endpoint has no namespace filter; filter client-side) export returned HTTP 200, total_rows=231 rewrote 5 memories from scenario38-src-fd1b46 -> scenario38-dst-fd1b46 bob imports the payload into scenario38-dst-fd1b46 on node-2 import returned HTTP 200 settle 6s for import + fanout verify row counts match on destination scenario38-dst-fd1b46 has 5 rows (expected 5) markers preserved in destination: 5/5
{
"agent_group": "hermes",
"checkpoint": "2026-04-23T20:43:43+00:00",
"diag_curl_body_head": "{\"count\":6,\"earliest_updated_at\":\"2026-04-23T20:44:17.337857441+00:00\",\"latest_updated_at\":\"2026-04-23T20:44:24.298240357+00:00\",\"limit\":500,\"memories\":[{\"access_count\":0,\"confidence\":1.0,\"content\":\"marker=delta-0-92e26d4d6a7f455a8c4a793ff317935e\",\"created_at\":\"2026-04-23T20:44:17.337857441+00:00\",\"",
"diag_curl_exit": 0,
"diag_curl_http_code": 200,
"diag_curl_stderr": "",
"diag_earliest_updated_at": "2026-04-23T20:44:17.337857441+00:00",
"diag_latest_updated_at": "2026-04-23T20:44:24.298240357+00:00",
"diag_node3_health_reachable": true,
"diag_updated_since": "2026-04-23T20:43:43+00:00",
"expected_markers": 6,
"markers_present": 6,
"namespace": "scenario39-delta-4d8df2",
"pass": true,
"reasons": [],
"rows_returned": 6,
"rows_returned_raw": 6,
"scenario": "39",
"skipped": false,
"tls_mode": "off"
}
checkpoint = 2026-04-23T20:43:43+00:00 suspending ai-memory on node-3 !! ssh timeout (30s): root@104.236.96.109 pgrep -f 'ai-memory serve' | xargs -r kill -STOP alice + bob write 6 rows while node-3 is out resuming ai-memory on node-3 settle 15s for process resume + federation catchup node-3 → node-1 health reachable: True (after 1 probes) node-3 asks node-1 /api/v1/sync/since?since=2026-04-23T20:43:43+00:00 curl exit=0 http_code=200 body_len=2898 stderr='' /sync/since raw=6 ns-filtered=6; 6/6 match our markers diag: updated_since=2026-04-23T20:43:43+00:00 earliest=2026-04-23T20:44:17.337857441+00:00 latest=2026-04-23T20:44:24.298240357+00:00
{
"agent_group": "hermes",
"bulk_http_code": "200",
"bulk_size": 500,
"namespace": "scenario40-bulk-2dc30b",
"pass": true,
"per_peer_count": {
"node_2": 500,
"node_3": 500,
"node_4": 500
},
"reasons": [],
"scenario": "40",
"skipped": false,
"tls_mode": "off"
}
constructing 500-row bulk payload staging bulk payload on node-1 /tmp, then POST /api/v1/memories/bulk bulk POST returned HTTP 200 settle 20s for bulk fanout across 3 peers + aggregator node-2 count=500 (expected 500) node-3 count=500 (expected 500) node-4 count=500 (expected 500)
{
"activity_namespace": "scenario41-activity-d9d265",
"agent_group": "hermes",
"pass": true,
"per_peer": {
"node_1": {
"counters_t0": 8,
"counters_t1": 8,
"regressed_keys": 0
},
"node_2": {
"counters_t0": 8,
"counters_t1": 8,
"regressed_keys": 0
},
"node_3": {
"counters_t0": 7,
"counters_t1": 7,
"regressed_keys": 0
}
},
"reasons": [],
"scenario": "41",
"skipped": false,
"tls_mode": "off"
}
scrape T0 node-1 T0 parsed 8 memory counters node-2 T0 parsed 8 memory counters node-3 T0 parsed 7 memory counters settle 5s for counter update scrape T1 node-1 T1 parsed 8 memory counters node-2 T1 parsed 8 memory counters node-3 T1 parsed 7 memory counters
{
"agent_group": "hermes",
"namespaces": [
"scenario42-abc471-0",
"scenario42-abc471-1",
"scenario42-abc471-2"
],
"pass": true,
"per_peer": {
"node_1": {
"scenario42-abc471-0": 2,
"scenario42-abc471-1": 2,
"scenario42-abc471-2": 2
},
"node_2": {
"scenario42-abc471-0": 2,
"scenario42-abc471-1": 2,
"scenario42-abc471-2": 2
},
"node_3": {
"scenario42-abc471-0": 2,
"scenario42-abc471-1": 2,
"scenario42-abc471-2": 2
},
"node_4": {
"scenario42-abc471-0": 2,
"scenario42-abc471-1": 2,
"scenario42-abc471-2": 2
}
},
"reasons": [],
"scenario": "42",
"skipped": false,
"tls_mode": "off"
}
alice writes into 3 distinct namespaces: ['scenario42-abc471-0', 'scenario42-abc471-1', 'scenario42-abc471-2']
settle 10s for namespace index fanout
node-1 sees 3/3 target namespaces, counts: {'scenario42-abc471-0': 2, 'scenario42-abc471-1': 2, 'scenario42-abc471-2': 2}
node-2 sees 3/3 target namespaces, counts: {'scenario42-abc471-0': 2, 'scenario42-abc471-1': 2, 'scenario42-abc471-2': 2}
node-3 sees 3/3 target namespaces, counts: {'scenario42-abc471-0': 2, 'scenario42-abc471-1': 2, 'scenario42-abc471-2': 2}
node-4 sees 3/3 target namespaces, counts: {'scenario42-abc471-0': 2, 'scenario42-abc471-1': 2, 'scenario42-abc471-2': 2}