feat: persistent agent memory + review-policy migration; openvang rebrand - #2374
Merged
Conversation
…rand Second slice of the local branch (runtime landed in the prior PR). Three coherent-but-interwoven changes that touch the same core files and so land together: - Persistent memory: benchmark/memory.py + source_memory.py + memory_coverage.py + ablation.py, threaded through agent/context.py, benchmark/runner.py, and scripts/run_eval.py, with a receipt-safe memory_commitment bound into benchmark/attestation.py (never binds raw recalled content). Memory-coverage and ablation protocols + their CLIs and tests. - Review-policy migration: REVIEW.md + CONTRIBUTING.md content moves into specs/009-agent-review; CODEOWNERS, the PR template, and the review scripts (agent/review.py, review_pr.py, benchmark_pr_policy.py, pr_reopen_policy.py) updated to match. - Identity: pyproject authors + URLs move gittensor-vanguard -> openvang, completing the org transfer. Blog posts are kept (only spec-driven-development.md is content-updated). Rebased onto the runtime PR's test; inherits its cryptography CI install.
github-actions
Bot
added
agent
|
Agent benchmark smoke check (offline stub — crash/shape only, not a reasoning-quality score)
This does not determine a perf:* label. That requires a maintainer-bot-run live benchmark comparison against both the public and private repo targets — see the benchmark policy. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Second (and final) slice of a large local branch — the entangled remainder after the OpenVang runtime (#2372) landed. Three coherent-but-interwoven changes that touch the same core files (
agent/context.py,benchmark/runner.py,scripts/run_eval.py,benchmark/attestation.py) and therefore can't be split further without hunk surgery:1. Persistent agent memory
benchmark/memory.py+source_memory.py+memory_coverage.py+ablation.py, threaded through the agent context / runner / run_eval path.memory_commitmentbound intobenchmark/attestation.py— schema/policy versions + snapshot/query/view digests only, never raw recalled content.2. Review-policy migration
REVIEW.md+CONTRIBUTING.mdcontent moves intospecs/009-agent-review;CODEOWNERS, the PR template, and the review scripts updated to match. Blog posts are kept (onlyspec-driven-development.mdis content-updated).3. Identity
pyprojectauthors + URLsgittensor-vanguard → openvang, completing the org transfer.Verification
Full suite green (rebased on current
test, inherits #2372'scryptographyCI install), ruff clean. Every commit author==committer==matedev01.One deliberate deferral (needs your call at review)
The local branch also rewrote
.dockerignoreinto a general product-image ignore, which removes the attested eval image's tight allowlist (docker/eval.Dockerfile's deliberately-small, auditable TCB that the Polaris attestation depends on). I did not take that rewrite here — I kepttest's eval allowlist. The product runtime and the attested eval image share one root.dockerignore; giving the product image its own ignore (or reconciling the two) is an architecture decision that belongs in its own change, not smuggled in here.Refs #2373.