CI / lint-and-test (push) Waiting to run

Details

Prove full live-safe Working Leo Telegram benchmark

- Run OE-01 through OE-05 live Telegram read-only benchmark coverage
- Retain full-suite scoring artifacts and no-mutation safety readback
- Broaden scorer tests for live Leo wording without weakening claim ceilings

Files:

2026-07-10 00:52:30 +02:00

20 KiB

Raw Blame History

Working Leo Current State - 2026-07-09

Generated UTC: 2026-07-09T21:15:00Z

Current Working Definition

A working Leo, for Cory/m3taversal's current expectation, is a Telegram-facing agent that remembers the current conversation, grounds KB answers in canonical Teleo Postgres rows, distinguishes proposed / approved / applied / not applied, and moves a specific Cory-approved KB change through a narrow review/apply path into canonical rows with before/after proof. Broad autonomous DB write authority is a feature change, not required for restoration.

What Is Now Proven On VPS

VPS gateway service is stable: leoclean-gateway.service is active/running, MainPID=3252143, NRestarts=0, WorkingDirectory=/home/teleo, User=teleo.
Telegram Bot/API readiness is proven for group Leo (-5146042086), and the live adapter accepts the Cory-working canary without posting.
Live Telegram-visible group canary passed through authenticated Chrome/Telegram Web:
- turn 1 posted at 21:31 Europe/Zurich and Leo replied LIVE_TG_TURN1_ACK
- turn 2 posted at 21:32 Europe/Zurich and Leo remembered marker WL-LIVE-TG-CORY-20260709
- turn 2 reply preserved the correction approved is not the same as applied
- turn 2 reply reported proposal 00957f6c-9883-4015-95a4-6b09367efb0e as applied
- turn 2 reply reported edge c167933e-d513-4f43-9335-d5d8aeb259f2 exists in public.claim_edges and the canonical row exists
Open-ended Cory-style Telegram canary passed:
- marker WL-LIVE-TG-CORY-OPEN-20260709
- Leo inferred the likely failure mode from "agents not working / KB same as last night" without exact proposal IDs
- Leo named kb_stage.kb_proposals, gateway/scheduler logs, and canonical public.* graph deltas as first checks
- Leo distinguished proposed, approved, and applied without staging or mutating the KB
Full live-safe open-ended Telegram suite now passes:
- OE-01 vague incident triage, OE-02 approved-vs-applied without exact IDs, OE-03 framework KB change, OE-04 runtime provenance, and OE-05 mixed action authorization all scored pass=True
- retained score shows coverage=full, passes=5, expected_prompt_count=5, and overclaim_detected=False for every prompt
- service stayed active/running after the suite with MainPID=3252143, NRestarts=0
- post-suite DB safety readback showed no marker rows in kb_stage or canonical public.* tables, and no claims/sources/edges/evidence rows created after the OE-02 start timestamp
- claim ceiling: this proves live read-only Telegram behavior, not production application of the prepared DB packets
Telegram-shaped handler canary passed in the same user/cwd context as the live service:
- authorized source: agent:main:telegram:group:-5146042086:9070919
- turn 1 ack passed
- turn 2 remembered marker and correction
- turn 3 confirmed proposal 00957f6c-9883-4015-95a4-6b09367efb0e is applied and edge c167933e-d513-4f43-9335-d5d8aeb259f2 exists in public.claim_edges
- no Telegram post, no live profile synthetic sessions, temp profile removed
Strict apply canary is proven: one strict add_edge proposal became a canonical public.claim_edges row, and read-only plan verification now passes with after-readback checks.
Guarded approval helper is installed on VPS at /opt/teleo-eval/workspaces/deploy-infra/scripts/approve_proposal.py, checksum c650b8b24b7a89a9e23859c04d850db441483d407fc81580e250ccad65a61d83, imports successfully, and rejects re-approval of the already-applied canary.
Chrome/Telegram Web is authenticated and read/write-proven for bot testing in the Leo group through Computer Use.
Live Telegram-to-DB staging is now proven:
- turn 3 posted at 22:06 Europe/Zurich and Leo replied LIVE_TG_TURN3_STAGED
- Leo created proposal 8dfedb3f-3aa4-4200-970f-4c0016f6869f in kb_stage.kb_proposals
- DB readback shows status=pending_review, proposal_type=add_edge, proposed_by_handle=leo-telegram-live, source_ref=telegram:leo:-5146042086:WL-LIVE-TG-CORY-20260709-T3
- canonical public tables stayed clean after the T3 timestamp: public.claims=0, public.sources=0, public.claim_edges=0, public.claim_evidence=0 new rows after 2026-07-09 20:06:30 UTC
- turn 4 posted at 22:08 Europe/Zurich and Leo replied LIVE_TG_TURN4_READBACK, remembered T3, and read back the staged proposal/status/source_ref without staging or applying anything else

What Is Still Not Proven

Old Cory-approved mapped rich proposals are still not fully directly applyable:
- 14fa5ecc-ac7a-41c1-807d-a2e85b936617
- ac036c9d-20a0-4ffe-881f-57d6b7bacf22
14fa5ecc normalization is blocked by missing canonical claim/source IDs. The proposal contains approved textual successor candidates, but the exact successor rows A/B/C/D do not already exist as canonical rows.
Current narrow apply worker supports add_edge, attach_evidence, and revise_strategy; it does not create new public.claims or public.sources rows. Applying 14fa5ecc as Cory likely expects requires either creating/identifying canonical rows first or explicitly widening the guarded apply contract.
Live approved rich proposal contract preview is now available:
- 14fa5ecc: 4 claim rows, 5 source rows, 17 edge fragments, 9 evidence fragments; not direct-apply-ready
- ac036c9d: 3 claim rows, 3 source rows, 6 edge fragments, 5 evidence fragments; not direct-apply-ready
Rollback-only rich proposal creation plan is now schema-rehearsed against VPS Postgres and left zero generated rows behind:
- 14fa5ecc: plans 4 claims, 9 sources, 11 claim_evidence rows, 3 supported claim_edges; blocks 14 fragments
- ac036c9d: plans 3 claims, 6 sources, 6 claim_evidence rows, 1 supported claim_edge; blocks 5 fragments
- the rehearsal exposed and fixed live constraint mismatches for claims.type and sources.source_type
- production application remains false until unsupported edge types/context mappings and design decisions are reviewed
Disposable staging DB commit proof is now complete for the supported subset:
- created teleo_rich_rehearsal_20260709 from live teleo via dump/restore
- committed both rich proposal transactions in the disposable DB
- staging readback showed 7 claims, 15 sources, 17 claim_evidence rows, and 4 claim_edges
- production generated-row counts stayed zero before and after
- staging DB and temp SQL/dump files were removed; final DB existence count was 0
Mapped rich proposal rehearsal is now complete for a larger supported subset:
- explicit overlay mapped supersedes_component_of -> supersedes, draws_from -> derives_from, revises_or_refines -> supersedes, and one existing capital-allocation target to claim c0000000-0000-0000-0000-00000000000c
- live rollback rehearsal passed on VPS Postgres and left zero generated rows
- disposable DB teleo_mapped_rehearsal_20260709 committed 7 claims, 15 sources, 17 evidence rows, and 11 claim_edges
- production generated-row counts stayed zero before and after
- clone and temp dump were removed; final DB existence count was 0
Production apply packet for the mapped supported subset is now generated and clone-proven:
- packet files live under outputs/rich-proposal-creation-plan-20260709/production-apply-packet-mapped-20260709/
- live preflight confirms both source proposals are approved, required existing claims are present, and production generated rows are 0/0/0/0
- live rollback rehearsal inserts the target rows in-transaction and rolls back cleanly; production remains 0/0/0/0
- disposable clone commit using the packet's authorized commit SQL produced 7 claims, 15 sources, 17 evidence rows, and 11 claim edges
- clone post-commit readback showed DB-side superseded_by pointers are set by generated supersedes edges for 7da4851c... and d0000000...
- packet delete rollback now clears those exact superseded_by pointers, deletes generated rows, and returns clone generated-row counts to 0/0/0/0
- production generated rows remain 0/0/0/0; disposable packet clone and temp dump were removed
Helmer 7 Powers approved proposal a64df080-8502-42e2-98f4-9bbdecb8da73 is now full-coverage packet-ready and ledger clone-proven:
- generated row contract is 1 reasoning tool, 5 claims, 13 sources, 7 claim_edges, and 17 claim_evidence rows
- canonical reasoning-tool description is sanitized so Teleo-specific self-positioning/application examples are deferred rather than canonically approved through the framework proposal
- live production preflight confirms the proposal is approved, required existing target claims are present, and generated rows are 0/0/0/0/0
- live production rollback rehearsal inserted the planned rows and proposal-ledger update inside transactions and rolled back cleanly
- disposable clone teleo_helmer_ledger_rehearsal_20260709 committed the packet and read back 1/5/13/7/17
- clone postflight read back proposal a64df080-8502-42e2-98f4-9bbdecb8da73 as applied with applied_by_handle=codex-packet-20260709
- clone delete rollback returned generated rows to 0/0/0/0/0 and restored the proposal ledger to approved
- production generated rows remain 0/0/0/0/0; disposable clone was dropped with final clone count 0
Cross-surface strategy anchor companion packet is now clone-proven:
- first clone attempt exposed a live schema constraint: strategy_node_anchors cannot target both claim_id and source_id
- packet was corrected so the anchor points only to Claim D, while approval provenance remains in Claim D evidence rows
- disposable DB teleo_cross_surface_rehearsal_20260709 applied the mapped base packet first (7 claims, 15 sources, 11 edges, 17 evidence rows)
- companion packet then updated anchor 47f128a0-36c5-4b60-97d4-48c32fa48bc7 from old claim d0000000-0000-0000-0000-000000000005 to Claim D e6ce4e36-15d2-53bc-8b74-dda0efc095c1 at weight 0.85
- cross-surface rollback restored the old anchor, base rollback returned generated rows to 0/0/0/0, production stayed unchanged, temp files were removed, and service remained active with MainPID=3252143, NRestarts=0
Rio strategy-context companion packet is now clone-proven:
- packet creates 2 Rio strategy nodes and 5 strategy anchors after the mapped base packet creates claims A/C/D
- disposable DB teleo_rio_strategy_rehearsal_20260710 applied the mapped base packet first (7 claims)
- Rio packet then inserted Rio capital deployment layer and Rio internet-finance capital product
- anchors connect Claim A to the capital layer, Claim C and Claim D to the product, the product to the capital layer, and the capital layer to the Back shared root
- Rio rollback deleted the 5 anchors and 2 nodes; base rollback returned generated base claims to 0; production stayed unchanged at 0/0/0; temp files and clone were removed; service remained active with MainPID=3252143, NRestarts=0
Governance/concept-map schema companion packet is now clone-proven:
- live schema had public.governance_gates but no claim-to-governance link table, and no first-class concept-map table
- packet creates minimal public.concept_maps, public.claim_concept_map_links, and public.claim_governance_gate_links tables in-transaction
- disposable DB teleo_governance_concept_rehearsal_20260710 applied the mapped base packet first (7 claims)
- governance/concept packet then inserted 1 concept map, 1 claim-concept link, and 2 claim-governance links
- rollback deleted generated rows and dropped the generated tables because they were empty; base rollback returned generated base claims to 0; production stayed unchanged with no generated tables; temp files and clone were removed; service remained active with MainPID=3252143, NRestarts=0
Integrated production-order packet rehearsal is now clone-proven:
- disposable DB teleo_integrated_rehearsal_20260710 applied the full prepared packet set in order: mapped base, cross-surface anchor, Rio strategy context, governance/concept schema, and Helmer 7 Powers
- clone apply footprint matched the integrated manifest: 12 claims, 28 sources, 18 claim edges, 34 claim evidence rows, 1 reasoning tool, 2 Rio strategy nodes, 5 Rio strategy anchors, 1 concept map, 1 concept link, 2 governance links, one Helmer ledger row marked applied, and one cross-surface anchor update
- reverse rollback restored generated base rows to 0/0/0/0, Helmer rows to 0/0/0/0/0, Helmer ledger to approved, cross-surface anchor to old claim d0000000-0000-0000-0000-000000000005 at weight 0.7, Rio rows to 0/0, and generated governance/concept schema to absent
- production readback after the clone proof stayed unchanged: generated base rows 0/0/0/0, Helmer rows 0/0/0/0/0, Helmer ledger approved, generated governance/concept schema absent, service active with MainPID=3252143, NRestarts=0
VPS/Hermes/DB identity and document-linking map is now documented:
- the supplied July 9 master/build and database-reference PDFs confirm the intended architecture is DB-first identity: canonical Postgres identity graph rows render to SOUL.md, and evidence depth remains on demand through teleo-kb
- direct SOUL.md edits are profile/runtime artifact edits, not canonical DB updates; they can drift or be overwritten unless represented in canonical rows first
- teleo-kb-apply-worker.service exists but is disabled/inactive and ships inert/report-only unless KB_APPLY_WORKER_ENABLED=1; no active KB_APPLY_RENDER_CMD hook or scheduled 24-hour SOUL renderer was proven
- current VPS schema does not contain the designed decision-matrix tables kb_stage.matrix_voters, kb_stage.proposal_votes, or kb_stage.proposal_decisions; current approval evidence lives in kb_stage.kb_proposals
- live proposal inspection showed pending/approved rows have source_ref pointers, but inspected source refs did not directly match public.sources; document proposals are real staging rows but still need guarded mapping into canonical sources/evidence/edges
GCP parity was rechecked and remains blocked on account authentication:
- billy@livingip.xyz is selected for project teleo-501523 but token refresh fails non-interactively
- billyattnmarket@gmail.com and valentin.meylan68@gmail.com can refresh tokens but do not have access to teleo-501523
- contact@attn.markets also fails token refresh
- active gcloud account was restored to billy@livingip.xyz

Current Evidence

Handler canary: outputs/telegram-gateway-handler-canary-20260709/telegram-gateway-handler-canary-current.md
Live Telegram canary: outputs/telegram-live-canary-20260709/telegram-live-canary-current.md
Live Telegram DB write canary: outputs/telegram-live-canary-20260709/telegram-live-db-write-canary-20260709.md
Retained live Telegram benchmark score: work/teleo-infra-main/docs/reports/leo-working-state-20260709/telegram-live-open-ended-benchmark-score-current.md
Full live-safe open-ended Telegram suite: work/teleo-infra-main/docs/reports/leo-working-state-20260709/telegram-live-open-ended-suite-current.md
Full live-safe open-ended Telegram suite score: work/teleo-infra-main/docs/reports/leo-working-state-20260709/telegram-live-open-ended-suite-score-current.md
Gateway adapter canary: outputs/telegram-gateway-canary-20260709/telegram-gateway-canary-current.md
KB apply plan: outputs/kb-apply-canary-20260709/kb-apply-canary-plan-current.md
KB apply execute: outputs/kb-apply-canary-20260709/kb-apply-canary-execute-current.md
Approval normalization: outputs/approved-proposal-normalization-20260709/approved-proposal-normalization-current.md
Claim/source contract preview: outputs/claim-source-contract-preview-20260709/claim-source-contract-preview-current.md
Rich proposal creation plan: outputs/rich-proposal-creation-plan-20260709/rich-proposal-creation-plan-current.md
Mapped rich proposal creation plan: outputs/rich-proposal-creation-plan-20260709/rich-proposal-creation-plan-mapped-current.md
Production apply packet for mapped subset: outputs/rich-proposal-creation-plan-20260709/production-apply-packet-mapped-20260709/production-apply-packet-current.md
Helmer 7 Powers apply packet: outputs/rich-proposal-creation-plan-20260709/helmer-7powers-20260709/production-apply-packet/production-apply-packet-current.md
Helmer 7 Powers ledger proof log: outputs/rich-proposal-creation-plan-20260709/helmer-7powers-20260709/production-apply-packet/ledger-clone-commit-delete-rollback-current.log
Cross-surface strategy anchor packet: outputs/rich-proposal-creation-plan-20260709/cross-surface-resolution-20260709/cross-surface-packet.json
Cross-surface strategy anchor clone proof log: outputs/rich-proposal-creation-plan-20260709/cross-surface-resolution-20260709/cross-surface-clone-rehearsal-current.log
Rio strategy-context packet: outputs/rich-proposal-creation-plan-20260709/rio-strategy-context-20260710/rio-strategy-packet.json
Rio strategy-context clone proof log: outputs/rich-proposal-creation-plan-20260709/rio-strategy-context-20260710/rio-strategy-clone-rehearsal-current.log
Governance/concept-map packet: outputs/rich-proposal-creation-plan-20260709/governance-concept-20260710/governance-concept-packet.json
Governance/concept-map clone proof log: outputs/rich-proposal-creation-plan-20260709/governance-concept-20260710/governance-concept-clone-rehearsal-current.log
Integrated packet manifest: outputs/rich-proposal-creation-plan-20260709/integrated-20260710/working-leo-integrated-packet.json
Integrated clone proof log: outputs/rich-proposal-creation-plan-20260709/integrated-20260710/working-leo-integrated-clone-rehearsal-current.log
VPS/Hermes/DB mapping: outputs/vps-hermes-db-mapping-20260709.md
Identity render / decision matrix / document artifact image: outputs/leo-db-state-22-identity-render-matrix-artifacts.svg
Execution plan: outputs/working-leo-execution-plan-20260709/working-leo-execution-plan-current.md
Cory expected outcomes: outputs/cory-expected-working-leo-outcomes-20260709.md
Current live proof image: outputs/leo-db-state-07-live-proof.svg
Rich proposal plan image: outputs/leo-db-state-08-rich-plan.svg
Staging commit proof image: outputs/leo-db-state-09-staging-commit.svg
Live Telegram staging proof image: outputs/leo-db-state-10-telegram-staged-write.svg
Mapped staging proof image: outputs/leo-db-state-11-mapped-staging.svg
Production packet proof image: outputs/leo-db-state-12-production-packet.svg
Helmer packet proof image: outputs/leo-db-state-14-helmer-packet.svg
Open-ended Cory-style canary image: outputs/leo-db-state-15-open-ended-canary.svg
Helmer ledger proof image: outputs/leo-db-state-16-helmer-ledger.svg
Cory-style benchmark split image: outputs/leo-db-state-17-cory-outcome-benchmark.svg
Leoclean DB/workspace image: outputs/leo-db-state-18-db-vs-workspace.svg
Claim/body/metadata schema image: outputs/leo-db-state-19-claims-body-metadata.svg
Cross-surface strategy anchor proof image: outputs/leo-db-state-20-cross-surface-anchor.svg
Rio strategy context proof image: outputs/leo-db-state-23-rio-strategy-context.svg
Governance/concept-map schema proof image: outputs/leo-db-state-24-governance-concept-schema.svg
Integrated apply rehearsal proof image: outputs/leo-db-state-25-integrated-apply-rehearsal.svg
GCP parity blocker: outputs/gcp-db-parity-current-blocker-20260709.json

Next Actions

Keep the live Telegram canary as the top-level regression proof for Cory-working memory + KB audit.
Do not auto-apply old freeform approvals.
Do not run production apply without explicit authorization. The mapped base, cross-surface anchor, Rio strategy, governance/concept, and Helmer packets compose in one clone and are ready as a packet set, but are not production-applied.
Keep VPS Leo runtime behavior unchanged while the DB apply path is stabilized.

20 KiB Raw Blame History