open-research-protocol 0.4.31 → 0.4.33
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/AGENT_INTEGRATION.md +4 -2
- package/CHANGELOG.md +35 -0
- package/cli/orp.py +990 -10
- package/docs/AGENT_LOOP.md +6 -1
- package/docs/ORP_AUTONOMY_PROJECT_COMPILATION_MODEL.md +3 -2
- package/docs/RESEARCH_COUNCIL.md +20 -0
- package/package.json +1 -1
package/AGENT_INTEGRATION.md
CHANGED
|
@@ -103,8 +103,10 @@ artifact paths (code/data/proofs/logs/papers).
|
|
|
103
103
|
- Treat **failed paths** as assets: record dead ends as a `Failed Path Record` with the blocking reason/counterexample and a
|
|
104
104
|
next hook.
|
|
105
105
|
- Resolve disputes by **verification or downgrade**, not argument.
|
|
106
|
-
- Run `orp hygiene --json` before long delegation, after material writeback, before
|
|
107
|
-
state grows unexpectedly.
|
|
106
|
+
- Run `orp hygiene --json` before long delegation, after material writeback, before remote side effects or unbudgeted paid
|
|
107
|
+
compute, and when dirty state grows unexpectedly.
|
|
108
|
+
- Do not hard-stop solely because an OpenAI research lane is paid; budgeted ORP research may run when `orp research` spend
|
|
109
|
+
preflight is within the configured daily cap.
|
|
108
110
|
- Stop long-running expansion while hygiene reports `dirty_unclassified`; classify, refresh generated surfaces, canonicalize
|
|
109
111
|
useful scratch, or write a blocker before continuing.
|
|
110
112
|
- Hygiene is non-destructive: never reset, checkout, or delete files merely to hide dirty state.
|
package/CHANGELOG.md
CHANGED
|
@@ -6,6 +6,41 @@ There was no prior in-repo changelog file, so the first formal entry starts
|
|
|
6
6
|
with the currently shipped `v0.4.4` release and summarizes the full release
|
|
7
7
|
delta reflected in this repo.
|
|
8
8
|
|
|
9
|
+
## v0.4.33 - 2026-04-30
|
|
10
|
+
|
|
11
|
+
This release teaches ORP to manage Codex's global instruction and startup hook
|
|
12
|
+
layer, and adds a modeled-checkoff frontier continuation for review-gated work.
|
|
13
|
+
|
|
14
|
+
### Added
|
|
15
|
+
|
|
16
|
+
- Added `orp agents codex audit` and `orp agents codex sync` to audit or
|
|
17
|
+
bootstrap Codex global `AGENTS.md`, `hooks.json`, the non-mutating
|
|
18
|
+
`SessionStart` hook script, and the `codex_hooks` feature flag.
|
|
19
|
+
- Added frontier modeled-checkoff detection and
|
|
20
|
+
`orp frontier modeled-checkoff activate` so review/checkoff blockers can be
|
|
21
|
+
converted into explicitly modeled packet-improvement work without claiming
|
|
22
|
+
real human, clinical, legal, regulatory, privacy, publication, or expert gate
|
|
23
|
+
clearance.
|
|
24
|
+
|
|
25
|
+
### Changed
|
|
26
|
+
|
|
27
|
+
- Frontier delegation preflight now reports modeled-checkoff eligibility when a
|
|
28
|
+
live frontier appears blocked on review or sign-off, while preserving existing
|
|
29
|
+
active additional work as the safer continuation when it exists.
|
|
30
|
+
|
|
31
|
+
## v0.4.32 - 2026-04-25
|
|
32
|
+
|
|
33
|
+
This release clarifies ORP's paid-work boundary so budgeted OpenAI research is
|
|
34
|
+
not treated as a hard stop solely because it uses paid API calls.
|
|
35
|
+
|
|
36
|
+
### Changed
|
|
37
|
+
|
|
38
|
+
- Built-in OpenAI research lanes now require a local spend policy, then use
|
|
39
|
+
spend preflight as the approval boundary for budgeted provider calls.
|
|
40
|
+
- Generated project context, AGENTS guidance, handoffs, and research docs now
|
|
41
|
+
distinguish budgeted ORP research from unbudgeted paid compute, purchases,
|
|
42
|
+
and cap-exceeded provider calls.
|
|
43
|
+
|
|
9
44
|
## v0.4.31 - 2026-04-25
|
|
10
45
|
|
|
11
46
|
This release refreshes ORP's OpenAI-backed research lanes and tightens
|