Skip to content

[codex] add episteme reasoning workflow#37

Open
GTrunSec wants to merge 59 commits into
mainfrom
episteme
Open

[codex] add episteme reasoning workflow#37
GTrunSec wants to merge 59 commits into
mainfrom
episteme

Conversation

@GTrunSec

Copy link
Copy Markdown
Contributor

Summary

  • Add the Episteme reasoning workflow surfaces for ontology review, Qianji scheduling, and live-model prompt audit artifacts.
  • Add structural IDF and source patch scaffolding so private ontology evidence can be reviewed before RDF mutation.
  • Extend Qianji activity wiring so episteme ontology reasoning is admitted through the workflow path instead of prompt-only handling.

Impact

This keeps ontology mutation behind review-only evidence and prepares the next slice: materializing source evidence into the Qianji LLM context for real private Episteme reasoning.

Validation

  • Pre-commit hook passed during commit, including all-features cargo check/clippy coverage.
  • Targeted package tests and clippy checks for xiuxian-wendao-episteme, xiuxian-wendao-studio, and xiuxian-qianji passed before the commit.
  • Root worktree was clean before push.

GTrunSec added 19 commits May 20, 2026 17:10
* feat(orchestrator): add polyglot Docling scheduler profile

* feat(orchestrator): own source-range worker policy

* perf(xiuxian-wendao-analyzer): auto fan out Docling OCR endpoints

* perf(xiuxian-wendao-studio): preserve source-range OCR gaps

* perf(xiuxian-wendao-studio): pack partial OCR cache gaps

* revert: drop partial OCR cache gap packing

* perf(orchestrator): tune source-range page target

* perf(xiuxian-wendao-studio): weight source-range OCR chunks

* feat(xiuxian-wendao-analyzer): route Docling OCR profiles

* perf(xiuxian-wendao-studio): route sub15 Docling OCR profiles

* feat(xiuxian-wendao-analyzer): add OCR2 OpenRouter provider path

* fix(xiuxian-wendao-studio): route OCR2 profiles through rendered pages

* feat(xiuxian-wendao-analyzer): add OCR2 remote tuning controls

* feat: gate OCR2 region recovery profile

* feat: add OCR2 scaffold recovery profile

* feat: tighten OCR2 recovery scheduling profile

* feat: guard OCR2 region recovery canaries

* feat: add Julia polyglot schedule bridge

* feat: generalize hosted VLM OCR recovery profiles

* refactor: remove hosted OCR alias surfaces

* feat: advance OCR2 and real search profiling

* perf: reduce real search proof query overhead

* fix(xiuxian-window): name session window API fields

* fix(ci): close harness and lint gates

* fix(ci): remove redundant must_use

* fix(ci): isolate macro env tests

* feat: advance OCR recovery and profile benchmarks

* feat: add backend text OCR fallback diagnostics

* feat: fail fast source-range OCR placeholders

* feat: gate verified empty backend OCR pages

* feat: fail on benchmark precision gate regressions

* feat: add wendaograph search strategy bridge

* feat: route graph strategy bridge to markdown sections

* feat: advance Docling recovery and repo projection flows

* feat: advance document recovery and llm routing

* fix: satisfy harness policy CI gates

* fix: close remaining CI lint gates

* fix: align builtin artifact helper contracts

* fix: repair daochang harness gate

* fix: close remaining CI lint failures

* refactor: harden polyglot orchestration gates

* refactor: migrate daochang bot runtime to lingchong

* fix: ci lint and package matrix

* fix: close ci feature drift

* fix: satisfy wendao lint gate

* refactor: enhance types harness

* feat: advance Wendao search strategy flow

* feat: advance search sync

* fix: enforce search strategy modularity gate

* fix: close ci package matrix

* fix: gate studio flight server support

* fix: compile db store benchmark

* fix: repair harness gates after policy update

* fix: tighten SearchStrategyFlow candidate discovery

* feat: land episteme and extraction workflow updates

* feat: land audio shard recovery workflow

Add the Rust-native workflow kernel, audio shard recovery planning, and Studio audio-shards route wiring.

Add analyzer hosted audio quality gates, diagnostic recovery helpers, benchmark reporting, and speech sidecar controls while preserving the existing Arrow contracts.

* feat: harden audio transcript ledger flow

* feat: add qianji control kernel

* refactor: split episteme source contracts

* feat: add episteme evidence cache bridge

* feat: add qianji control storage adapters

* test: avoid unwraps in qianji valkey tests

* chore: update wendao episteme contracts

* feat: project qianji workflow traces to control events

* fix: preserve wendao evidence gates

* feat: land qianji control and episteme surfaces

* feat: expose org sdd status recovery

* feat: expose qianji control history

* feat: expose qianji control view

* ci: disable recursive submodule checkout

* feat: expose qianji control step

* fix: align bounded work sql tests with duckdb

* fix: remove episteme test unwrap assertions

* fix: split orgize archive runtime test

* ci: restore authenticated recursive submodules

* fix: remove duckdb local relation test expects

* fix: satisfy strict test clippy lints

* feat: expose qianji control activity

* feat: add qianji control decision cli

* feat: expose qianji control timer

* feat: add qianji control signal

* feat: add qianji control state query

* feat: add qianji control heartbeat

* feat: add qianji worker heartbeat journal

* feat: add qianji step queue journal

* feat: add qianji retry recovery applier

* feat: add qianji timer recovery applier

* feat: add qianji lease reclaim applier

* feat: add qianji recovery attempt journal

* feat: add qianji bounded recovery loop

* feat: add orgize sdd graph checks

* refactor: split orgize read model modules

* refactor: split orgize read model store

* feat: add qianji hot state snapshots

* refactor: split orgize command modules

* feat: expose qianji hot state snapshot

* feat: add qianji recovery apply CLI

* feat: add qianji activity queue projection

* feat: add qianji activity start CLI

* feat: add qianji activity finish CLI

* feat: summarize qianji activity lifecycle queue

* fix: clarify qianji in-flight activity summary

* feat: add qianji lease inspection CLI

* fix: keep cargo wrapper rustc reachable

* fix: use full rust toolchain for clippy hook

* feat: advance qianji lease inventory and validation surfaces

* chore: stabilize qianji and wendao gates

* ci: harden rust toolchain wrapper

* ci: bootstrap macos rustup

* ci: stabilize rust test lane

* ci: harden valkey and python binding checks

* docs: document ontology quality proof discovery
@GTrunSec GTrunSec force-pushed the episteme branch 3 times, most recently from a6d82a2 to 6f45773 Compare May 29, 2026 03:16
@GTrunSec GTrunSec marked this pull request as ready for review May 30, 2026 15:18
@chatgpt-codex-connector

Copy link
Copy Markdown

💡 Codex Review

PORT="${WENDAO_AI_PORT:-9518}"

P1 Badge Use a distinct default port for wendao-ai

In the default process-compose configuration I checked, nix/modules/process.nix now starts both wendao-frontend and wendao-ai, while the existing frontend launch script already defaults to port 9518. This new default makes the normal runtime start two web dev servers on the same listener, so one service will fail to bind or the wendao-ai healthcheck can probe the frontend instead unless every operator overrides WENDAO_AI_PORT.


is_frontend_healthy = is_runtime_web_healthy

P2 Badge Preserve the frontend health helper signature

For Python callers that import the compatibility helper directly, this alias changes is_frontend_healthy(...) from the previous frontend-specific signature to is_runtime_web_healthy(...), which requires a service_name keyword. Existing tests or external scripts that pass the old arguments now fail with TypeError before performing any healthcheck; wrap the shared helper with service_name="wendao-frontend" instead of aliasing it directly.

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

GTrunSec added 30 commits June 1, 2026 09:43
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant