[engram-parser] GH#10 — safetensors: coordinate ecosystem / sibling parser extraction from corinth-canal (separate from GGUF #7; one-way copy, no dep on corinth-canal)

⚠ SUPERSEDED PLAN — read this first (2026-08-15)

Everything below describing a dedicated safetensors-parser crate and an "engram-parser charter remains GGUF-only" is reversed. Safetensors is ported into engram-parser behind an off-by-default safetensors cargo feature. No new repo. The body is preserved unedited below for history; the current decision and its rationale are in the comment thread and in engram-parser's README.md § *Origin / modularization (*#10).

Still true and unchanged: one-way copy from inspiration; no dep on corinth-canal and no corinth dep on engram-parser; corinth keeps an unmodified reference copy; config/map are not ported.

Also stale below: "Limen-Neural org" (both repos moved to rmems), "toward frozen" (module stays reference permanently), and "LIM-88" (archived/Duplicate — this issue is the live tracker).

The acceptance criteria below have been edited in place — ACs 5, 6, 9 and 10 needed their text changed, not just their boxes ticked. See the comment thread.

Source

Synced / linked to GitHub for 1:1 mirror (2026-07-26).

GitHub: [engram-parser] GH#10 — safetensors: coordinate ecosystem / sibling parser extraction from corinth-canal (separate from GGUF #7; one-way copy, no dep on corinth-canal) #10 — OPEN
Repo: Limen-Neural/engram-parser GH#10
Match confidence: 1.15

Summary

Track the separate extraction (one-way copy of implementation as inspiration) of the reusable Safetensors header inspection, deterministic manifest generation, and MoE router/expert candidate discovery logic from the experimental reference implementation in rmems/corinth-canal (parallel to the GGUF work in this repo's #7).

This is the separate dedicated issue in engram-parser (per user request) for the Safetensors case. Companion/parallel to GGUF #7 / corinth#115. See corinth-canal#116 (the supporting/bootstrap issue in the source experimental repo) and Limen-Neural/cortex-tensor#9 (coord).

Important clarification (per modularization intent and prior reviews): This is one-way extraction/copying. We are not making corinth-canal depend on any new or target crate. corinth-canal remains self-contained as the experimental end-to-end reference (per its own PROMOTION_RULES.md: after "frozen", it keeps an unmodified reference copy of the logic). "we are not making corinth-canal dep. We are simply copying and implementing code from inspiration." "corinth-canal repo more of my experimental repo. Limen-Neural organization repos are more about modular reusable libraries. Ideally I am trying to extract useful code from corinth-canal repo."

Context

The Safetensors support in corinth-canal (src/moe/safetensors.rs + safetensors/discovery.rs) is a self-contained pure-Rust implementation (header parse without the official safetensors crate for the header/inspect path, sharded/HF index/dir support, inspect_safetensors_checkpoint / write_safetensors_manifest, SafetensorsManifest with tensor records (name/dtype/shape/byte_size/offsets/shard) + metadata + checkpoint source, MoE candidate labeling via name/shape heuristics for routers/experts/groups (classify_tensor, discover_candidates, SafetensorsRouterCandidate, SafetensorsExpertGroup, layout family detection e.g. "generic_moe", "deepseek_v3_family", "phimoe" etc.), plus MappedSafetensorsCheckpoint for loading tensor bytes on demand).
It is used for local model onboarding, examples/safetensors_manifest.rs, verification checklist, and dual-backend (with GGUF) in the Router/Checkpoint for experiments.
Currently marked "reference" (implied under moe/checkpoint in MODULE_STATUS.md targeting rmems-moe).
No Limen-Neural crate yet owns the reusable Safetensors parser + MoE discovery surface (unlike GGUF → engram-parser). Enram-parser remains GGUF-only per its charter (this issue is for ecosystem/sibling coordination of the parallel extraction).
Phase 1 (GGUF) issues already created: engram#7 (primary in this repo), corinth#115. corinth#115 non-goals noted "Do not move Safetensors handling (separate module) unless trivial" — this work (tracked via the separate corinth#116 + this engram issue) makes it non-trivial.
Parallel extractions (cortex-tensor from corinth) and the overall LIM-9 boundary matrix favor focused reusable libs under Limen-Neural. corinth-canal#116 exists as the source-side supporting issue (with note that primary work for a new crate will follow); this is the separate tracking issue here in engram-parser (the GGUF home) for the Safetensors extraction story.
Supporting corinth-canal#116 and cortex#9 exist; this creates the requested separate issue in engram-parser.

Related

rmems/corinth-canal#116 (supporting extraction issue in source repo; cross-link + update its comment)
Phase 1 GGUF siblings: Limen-Neural/engram-parser#7 (this repo's GGUF primary; this issue is the Safetensors separate parallel), rmems/corinth-canal#115 (migration side), moe/gguf: coordinate with engram-parser extraction from corinth-canal; avoid overlapping parser work cortex-tensor#8 (GGUF coord) + #9 (Safetensors coord), Linear LIM-88 (extend), beads
corinth docs/PROMOTION_RULES.md (frozen = reference copy kept unmodified in corinth), MODULE_STATUS.md (update safetensors layer), MODEL_SOURCE_VERIFICATION_CHECKLIST.md, configs/safetensors, examples/safetensors_manifest.rs
LIM-9 (org boundary matrix)
engram-parser README / scope (add ecosystem note for sibling parser extraction from corinth-canal)

Goal

Ensure the Safetensors extraction from corinth-canal (as inspiration, one-way copy, no dep) is tracked with a dedicated separate GitHub issue in this repo (engram-parser), parallel to #7. This provides visibility in the parser-related issues here, cross-links the full story (corinth source #116, cortex coord #9, future reusable crate work, LIM-9), and follows the precedent of explicit GitHub issues for such modularization.

(The actual reusable crate implementation can be a new Limen-Neural/safetensors-parser or placed per future decision; this issue tracks the extraction effort and ecosystem from engram's perspective.)

Non-goals

Do not add a runtime or build dependency from corinth-canal's Cargo.toml or code paths.
Do not change corinth-canal's Router/CheckpointBackend/Safetensors experiment paths, outputs, or examples (they continue using the local reference copy).
Do not move corinth-specific extensions.
This issue does not change engram-parser's GGUF-only implementation surface (engram charter preserved; this is coordination/ecosystem).
No full matmul/dequant/CUDA (pure parser/extractor like the GGUF side).
Do not create new repo in this plan (per user revision).

Acceptance criteria

Validation

Manual review of new issue, all cross-comments, README, docs updates.
Links render; language matches prior reviews ("no dep...", "copying and implementing code from inspiration").
(Later, when port happens) New reusable surface produces equivalent manifests/candidates to corinth on test cases from corinth configs.

Suggested branch

docs/safetensors-sibling-extraction (or feat/safetensors-ecosystem if any engram updates).

References

Primary plan: this engram-parser session plan.md
corinth-canal#116 (bootstrap/supporting in source; the "extract ... from corinth-canal")
Phase 1: engram-parser#7, corinth#115, cortex#8/#9, Linear LIM-88 + beads
corinth src/moe/safetensors.rs + discovery.rs + examples + configs + docs/PROMOTION_RULES.md + MODULE_STATUS.md
LIM-9

Created as the separate GitHub issue in engram-parser for the Safetensors extraction from corinth-canal (parallel to #7; per user request for the issue here, one-way copy, no dep on corinth-canal, following modularization pattern from LIM-9 and GGUF precedent).

Siblings (cross-linked/updated together):

rmems/corinth-canal#116 (supporting; already existed, link + comment updated)
Limen-Neural/cortex-tensor#9 (coord; link + comment updated)
engram-parser#7 (GGUF primary in this repo; comment updated)
Linear LIM-88, beads in corinth

Decision reversed: safetensors lands in engram-parser behind a cargo feature

Charter reversal (2026-08). This issue, the engram-parser README, corinth-canal docs/MODULE_STATUS.md, and cortex-tensor#9 all previously stated that the extractable safetensors surface would land in a new dedicated safetensors-parser crate, and that "engram-parser charter remains GGUF-only." That is reversed. The extractable set — manifest, discovery, json, paths, validate from corinth's src/moe/safetensors/ — is ported into engram-parser behind an off-by-default safetensors cargo feature. No safetensors-parser repo was or will be created.

Why: the charter was never "GGUF". It is zero-dependency checkpoint deserialization plus MoE raw-weight extraction — and safetensors header inspection is exactly that shape: header-only parse, deterministic manifest, name/shape MoE candidate discovery, raw bytes out, no math, no mmap. A third crate would have duplicated this crate's error type, dtype model, MSRV policy, CI, Docker, and release plumbing to host ~1.4k lines that share every one of its invariants. Consumers had already converged on this home without being asked: rmems/hybrid-fusion#27 states "Concrete memmap / header parse / payload extract → engram-parser (same rule as GGUF)". A cargo feature buys the same isolation a separate crate would have: default builds stay byte-for-byte GGUF-only, and the zero-dependency guarantee holds in every feature combination — [dependencies] stays empty, and the upstream safetensors and serde_json crates remain forbidden dependencies.

Unchanged, and still governing: this is a one-way copy from inspiration. engram-parser takes no dependency on corinth-canal, and corinth-canal takes no dependency on engram-parser for safetensors. corinth-canal keeps its unmodified reference copy of src/moe/safetensors/ and keeps using it in its Router / CheckpointBackend experiment paths. The corinth-specific config (HF config.json) and map (mmap load / token extract) modules are not ported.

What this changes in this issue

The body is preserved unedited under a superseded banner rather than rewritten, so the original intent stays legible. Four acceptance criteria needed their text changed, not just their boxes ticked:

AC 9 is directly contradicted. "engram stays GGUF-only" is the exact thing being reversed. Its other clause, "Limen-Neural org for reusable libs", is also stale — engram-parser and cortex-tensor were transferred to rmems. The two surviving clauses (one-way copy, no dep) are preserved verbatim everywhere.
AC 5's "MODULE_STATUS for safetensors layer toward frozen" is dropped. The module stays at reference permanently: corinth's ladder defines frozen as "the modular crate becomes the source of truth", and that is not happening — both copies stay live and are expected to diverge. corinth's docs/PROMOTION_RULES.md gains a One-way extractions section documenting the departure.
AC 6's "Linear LIM-88" is stale — LIM-88 is archived with status Duplicate. The live tracker is RM-344.
AC 10 has no committable target. There is no plan.md at repo root, and docs/superpowers/ is gitignored by design (.gitignore:15, documented in REVIEW.md §7). The committed record is instead README.md § Origin / modularization ([engram-parser] GH#10 — safetensors: coordinate ecosystem / sibling parser extraction from corinth-canal (separate from GGUF #7; one-way copy, no dep on corinth-canal) #10) + CHANGELOG.md. Noting this so it is not re-litigated later as an unmet AC.

AC 4 is reinterpreted: it asks for an "ecosystem/sibling note", but there is no sibling. It becomes an in-crate charter note.

Most cross-linking ACs were already satisfied

corinth#116, cortex-tensor#9, and engram#7 already carried the #10 link and the "no dep on corinth-canal" language before this pass. What was actually missing was the reversal.

Not blocked by #45

This surface is header-only — no mmap, no dequant. It does not wait on #45, which blocks the GGUF consumer story in corinth-canal#115. Worth stating explicitly since RM-344 is already relatedTo the mmap issues in Linear.

In flight


engram-parser docs	#57
corinth-canal docs	rmems/corinth-canal#156
the code port	follow-up PR, branch `feat/safetensors-10-port`

Still stale elsewhere

rmems/cortex-tensor README.md:79/:113 and src/moe/mod.rs:33, plus the engram-parser and cortex-tensor GitHub wikis, still describe the safetensors-parser crate. Tracked separately.

(Housekeeping: the bead id raulmc-yqj cited in comments on corinth#116 and engram#7 does not exist — the real bead is corinth-canal-le53qh.)

docs: address PR #57 bot review — planned-not-shipped, scope no-dep a…

docs(module-status): retarget safetensors to engram-parser feature; k…

docs(promotion-rules): document one-way extractions departing from th…

docs(moe): note safetensors extraction target in module docs

docs(moe): retarget safetensors inspect half to engram-parser feature…

docs: reverse #10 — safetensors ships in-crate behind a cargo feature…

@rmems

2026-08-22 architecture clarification

Current intent is stronger than the historical "copy forever, no dependency in either direction" wording preserved in this issue body.

engram-parser is the canonical reusable home for Safetensors header deserialization, deterministic manifests, and MoE router/expert discovery behind its off-by-default safetensors feature.
engram-parser remains zero-dependency: this means supporting/parsing the Safetensors format; it does not mean adding the upstream Rust safetensors crate as a dependency.
engram-parser must never depend on corinth-canal.
Corinth may retain its local Safetensors reference implementation while the reusable feature is incomplete, but under rmems/corinth-canal#161 the long-term architecture should re-evaluate Corinth consuming engram-parser once the required reusable surface, parity, mmap/payload boundaries, and performance gates are adequate. Permanent duplication is no longer the desired default.
No separate safetensors-parser repository should be created.

This lane belongs to Corinth's v0.3.0 extraction program (rmems/corinth-canal#161, milestone #11).

A focused README cleanup issue is being created so the README stops describing Safetensors as "copy, never a dependency" while still being precise that the feature is not shipped until implementation lands.

Planning attribution: GPT-5.6 Sol (ChatGPT/OpenAI) with @rmems, 2026-08-22.

@coderabbitai plan

Coding Plan

Summary

Treat GH#10 as a coordination and documentation ticket. Add no Rust code, no safetensors feature, and no dependency, per the ticket non-goals.

Audit and align the two committable in-repo records, README.md and CHANGELOG.md, so both carry the four required clarifications, the reversal statement, the source/exclusion boundaries, and the cross-links.

Confirm REVIEW.md does not contradict the aligned documentation. Leave the intentional docs/superpowers/ gitignore untouched.

Route all cross-repository and issue-tracker actions to the follow-up section, because they are not committable in this repository.

Follow-up actions outside this repository (not committable here)

The following acceptance criteria cannot be satisfied by changes in the engram-parser codebase. Record them for the owner or a coordination agent to execute in the named external systems. Use the same clarification language in every place: one-way copy from inspiration; rmems org for reusable libraries; no dependency on corinth-canal in either direction; engram-parser owns safetensors behind an off-by-default safetensors cargo feature.

Create or confirm the dedicated GitHub issue Limen-Neural/engram-parser#10 with full cross-links and the "no dep on corinth-canal" language.
Add or update cross-link comments on corinth-canal#116, cortex-tensor#9, and this repo's #7.
Update corinth-canal docs: retarget docs/MODULE_STATUS.md and manifests/proven_components.toml to rmems/engram-parser (feature safetensors), keep status reference (not "toward frozen"), and add a docs/PROMOTION_RULES.md "One-way extractions" section. Add explicit "no dep" language on corinth-canal#115 and corinth-canal#116.
Update Linear rmems/engram-parser#10 with the reversal. Note that LIM-88 is archived/Duplicate and rmems/engram-parser#10 is its live successor.
Create or update beads in corinth-canal for the separate Safetensors extraction.
Acknowledge coordination with a comment in cortex-tensor#7 or #9.

Design Choices

Design Choice 1: Code, feature-flag, or module scaffolding requirement

Options Considered:

Scaffold src/safetensors/ and add a safetensors = [] feature now.
Treat the ticket as documentation and coordination only, with no code.

Chosen Option: 2

Rationale: The ticket states that it does not change engram-parser's GGUF-only implementation surface. It states that no code has landed yet and that this entry records the decision only. The last acceptance criterion names README.md and CHANGELOG.md as the only committable record.

Design Choice 2: Scope of in-repo work given existing documentation

Options Considered:

Assume the docs are complete and make no changes.
Audit the existing docs against the ticket's explicit clarification checklist and refine any gaps.

Chosen Option: 2

Rationale: The acceptance criteria list four specific clarifications that every place must carry. The safe action is a targeted audit-and-align pass, not an assumption of completeness.

💡 User Tips

Regenerate the plan with different choices with @coderabbitai <feedback>.

Implementation Steps

In-repo documentation alignment

This phase makes the two committable records in this repository consistent with the superseded-plan decision. The goal is to guarantee that every required clarification appears, that cross-links render, and that the wording matches prior reviews. This phase is the full in-repo scope of the ticket. It adds no source code, no feature flag, and no dependency.

Task 1: Audit and align the README ecosystem/sibling note

Confirm that the safetensors sections carry all four ticket clarifications. Correct any gaps.

Review README.md sections "Planned / in flight — Safetensors", "Origin / modularization — Safetensors (#10)", the "Scope / boundaries" table, and the "Ecosystem / promotion model".
Verify the four required clarifications are present and explicit: (1) one-way copy from inspiration; (2) rmems org for reusable libraries; (3) no dependency on corinth-canal in either direction; (4) engram-parser owns safetensors behind an off-by-default safetensors cargo feature.
Verify the reversal statement remains: the previous "permanently GGUF-only" and "separate safetensors-parser repository" plan is superseded. No separate parser repository is created.
Verify the source-boundary text names corinth-canal/src/moe/safetensors/ as the promote-from surface. It must list the "do not promote" exclusions (Corinth config.json adapter policy, runtime/payload mmap, GPU registration/dequantization, SAAQ orchestration, machine-local config).
Verify the zero-dependency charter statement stays: no upstream safetensors crate, no serde_json, [dependencies] stays empty in every feature combination.
Verify cross-reference links to engram-parser#7, engram-parser#10, engram-parser#61, corinth-canal#115, corinth-canal#116, corinth-canal#161, and cortex-tensor#9 are present and render.
If any clarification, link, or exclusion is missing or inconsistent, edit the prose to add it. Keep the existing section structure and heading names.

Task 2: Align the CHANGELOG Unreleased entry

Confirm the [Unreleased] entry records the decision with the required clarifications and issue references.

Review the [Unreleased] → ### Changed "Charter reversal (#10)" entry in CHANGELOG.md.
Verify the entry states: safetensors ships inside this crate behind an off-by-default safetensors cargo feature; not in a separate safetensors-parser crate; the zero-dependency guarantee is unchanged; the extraction is a one-way copy with no dependency in either direction; no code has landed yet.
Verify the entry cites the relevant issues and cross-repo docs already referenced (#10, #45, corinth-canal docs/MODULE_STATUS.md, cortex-tensor#9).
Keep the Keep-a-Changelog format: bolded short label, issue references, bullet under ## [Unreleased].
If the entry omits any of the four ticket clarifications, extend the existing bullet. Do not add a duplicate entry.

Task 3: Confirm REVIEW.md consistency

Ensure REVIEW.md does not contradict the aligned README and CHANGELOG.

Review REVIEW.md charter language and its §7 .gitignore note.
Confirm REVIEW.md describes the charter as "GGUF today, plus safetensors headers once #10 lands", consistent with README.md.
Confirm no remaining unqualified "GGUF-only" charter claim exists that contradicts the reversal.
If a contradicting claim exists, update the prose to match the README wording. If none exists, make no change.
Do not change the §7 note about docs/superpowers/. It is intentional background, not an actionable item.

🤖 Prompt for AI agents

Implement Phase 1: In-repo documentation alignment. This phase is
documentation-only. Add no source code, no feature flag, and no dependency.

Task 1 — Audit and align the README ecosystem/sibling note:
- Review `README.md` sections "Planned / in flight — Safetensors", "Origin /
modularization — Safetensors (`#10`)", the "Scope / boundaries" table, and the
"Ecosystem / promotion model".
- Verify these four clarifications are present and explicit: (1) one-way copy
from inspiration; (2) `rmems` org for reusable libraries; (3) no dependency on
`corinth-canal` in either direction; (4) `engram-parser` owns safetensors behind
an off-by-default `safetensors` cargo feature.
- Verify the reversal statement is present: the previous "permanently GGUF-only"
and "separate `safetensors-parser` repository" plan is superseded, and no
separate parser repository is created.
- Verify the source-boundary text names `corinth-canal/src/moe/safetensors/` as
the promote-from surface, and lists the "do not promote" exclusions: Corinth
`config.json` adapter policy, runtime/payload mmap, GPU
registration/dequantization, SAAQ orchestration, machine-local config.
- Verify the zero-dependency charter statement is present: no upstream
`safetensors` crate, no `serde_json`, `[dependencies]` stays empty in every
feature combination.
- Verify cross-reference links to `engram-parser#7`, `engram-parser#10`,
`engram-parser#61`, `corinth-canal#115`, `corinth-canal#116`,
`corinth-canal#161`, and `cortex-tensor#9` are present and render correctly.
- Edit the prose to add any missing clarification, link, or exclusion. Keep the
existing section structure and heading names unchanged.

Task 2 — Align the CHANGELOG Unreleased entry:
- Review the `[Unreleased] → ### Changed` "Charter reversal (`#10`)" entry in
`CHANGELOG.md`.
- Verify the entry states: safetensors ships inside this crate behind an
off-by-default `safetensors` cargo feature (not a separate `safetensors-parser`
crate); the zero-dependency guarantee is unchanged; the extraction is a one-way
copy with no dependency in either direction; no code has landed yet.
- Verify the entry cites `#10`, `#45`, `corinth-canal docs/MODULE_STATUS.md`,
and `cortex-tensor#9`.
- Keep the Keep-a-Changelog format: bolded short label, issue references, bullet
under `## [Unreleased]`.
- If any of the four ticket clarifications is missing, extend the existing
bullet instead of adding a duplicate entry.

Task 3 — Confirm REVIEW.md consistency:
- Review the `REVIEW.md` charter language and its §7 `.gitignore` note.
- Confirm `REVIEW.md` describes the charter as "GGUF today, plus safetensors
headers once `#10` lands", matching `README.md`.
- Confirm no unqualified "GGUF-only" charter claim remains that contradicts the
reversal.
- If a contradicting claim exists, update the prose to match the README wording;
otherwise make no change.
- Leave the §7 note about `docs/superpowers/` unchanged, since it is intentional
background.

Research

engram-parser is a pure-Rust, zero-dependency crate. src/lib.rs declares error, gguf, and moe modules. No #[cfg(feature)] gating exists today. Cargo.toml has empty [dependencies] and [features] default = []. No safetensors code, tests, or examples exist.

README.md already carries a "Planned / in flight — Safetensors" section, an "Origin / modularization — Safetensors (#10)" section, a scope/boundaries table, and an ecosystem/promotion model.

CHANGELOG.md already carries a single [Unreleased] → ### Changed "Charter reversal (#10)" entry.

REVIEW.md §7 documents the intentional docs/superpowers/ gitignore.

The in-repo work for this ticket is to audit and align existing documentation against the ticket's required clarifications. It does not introduce new concepts.

🚀 Next Steps

🤖 All AI agent prompts combined

Implement Phase 1: In-repo documentation alignment. This phase is
documentation-only. Add no source code, no feature flag, and no dependency.

Task 1 — Audit and align the README ecosystem/sibling note:
- Review `README.md` sections "Planned / in flight — Safetensors", "Origin /
modularization — Safetensors (`#10`)", the "Scope / boundaries" table, and the
"Ecosystem / promotion model".
- Verify these four clarifications are present and explicit: (1) one-way copy
from inspiration; (2) `rmems` org for reusable libraries; (3) no dependency on
`corinth-canal` in either direction; (4) `engram-parser` owns safetensors behind
an off-by-default `safetensors` cargo feature.
- Verify the reversal statement is present: the previous "permanently GGUF-only"
and "separate `safetensors-parser` repository" plan is superseded, and no
separate parser repository is created.
- Verify the source-boundary text names `corinth-canal/src/moe/safetensors/` as
the promote-from surface, and lists the "do not promote" exclusions: Corinth
`config.json` adapter policy, runtime/payload mmap, GPU
registration/dequantization, SAAQ orchestration, machine-local config.
- Verify the zero-dependency charter statement is present: no upstream
`safetensors` crate, no `serde_json`, `[dependencies]` stays empty in every
feature combination.
- Verify cross-reference links to `engram-parser#7`, `engram-parser#10`,
`engram-parser#61`, `corinth-canal#115`, `corinth-canal#116`,
`corinth-canal#161`, and `cortex-tensor#9` are present and render correctly.
- Edit the prose to add any missing clarification, link, or exclusion. Keep the
existing section structure and heading names unchanged.

Task 2 — Align the CHANGELOG Unreleased entry:
- Review the `[Unreleased] → ### Changed` "Charter reversal (`#10`)" entry in
`CHANGELOG.md`.
- Verify the entry states: safetensors ships inside this crate behind an
off-by-default `safetensors` cargo feature (not a separate `safetensors-parser`
crate); the zero-dependency guarantee is unchanged; the extraction is a one-way
copy with no dependency in either direction; no code has landed yet.
- Verify the entry cites `#10`, `#45`, `corinth-canal docs/MODULE_STATUS.md`,
and `cortex-tensor#9`.
- Keep the Keep-a-Changelog format: bolded short label, issue references, bullet
under `## [Unreleased]`.
- If any of the four ticket clarifications is missing, extend the existing
bullet instead of adding a duplicate entry.

Task 3 — Confirm REVIEW.md consistency:
- Review the `REVIEW.md` charter language and its §7 `.gitignore` note.
- Confirm `REVIEW.md` describes the charter as "GGUF today, plus safetensors
headers once `#10` lands", matching `README.md`.
- Confirm no unqualified "GGUF-only" charter claim remains that contradicts the
reversal.
- If a contradicting claim exists, update the prose to match the README wording;
otherwise make no change.
- Leave the §7 note about `docs/superpowers/` unchanged, since it is intentional
background.

💡 Iterate on the plan with: `@coderabbitai` <feedback>

Example Feedback
- `@coderabbitai` You can skip phase 3. Add a simple unit test case for phase 2.
- `@coderabbitai` For design choice 1 go ahead with option 3 and replan.

feat: safetensors discovery and optional mmap GGUF dequant (#73)

Navigation Menu

Repository navigation

[engram-parser] GH#10 — safetensors: coordinate ecosystem / sibling parser extraction from corinth-canal (separate from GGUF #7; one-way copy, no dep on corinth-canal) #10

⚠ SUPERSEDED PLAN — read this first (2026-08-15)

Source

Summary

Context

Goal

Non-goals

Acceptance criteria

Validation

Suggested branch

References

Decision reversed: safetensors lands in engram-parser behind a cargo feature

What this changes in this issue

Most cross-linking ACs were already satisfied

Not blocked by #45

In flight

Still stale elsewhere

5 remaining items

2026-08-22 architecture clarification

Coding Plan

Summary

Follow-up actions outside this repository (not committable here)

Implementation Steps

In-repo documentation alignment

🚀 Next Steps

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

[engram-parser] GH#10 — safetensors: coordinate ecosystem / sibling parser extraction from corinth-canal (separate from GGUF #7; one-way copy, no dep on corinth-canal) #10

Description

⚠ SUPERSEDED PLAN — read this first (2026-08-15)

Source

Summary

Context

Goal

Non-goals

Acceptance criteria

Validation

Suggested branch

References

Activity

rmems commented on Aug 15, 2026

Decision reversed: safetensors lands in engram-parser behind a cargo feature

What this changes in this issue

Most cross-linking ACs were already satisfied

Not blocked by #45

In flight

Still stale elsewhere

5 remaining items

rmems commented on Aug 22, 2026

2026-08-22 architecture clarification

rmems commented on Aug 23, 2026

coderabbitai commented on Aug 23, 2026

Coding Plan

Summary

Follow-up actions outside this repository (not committable here)

Implementation Steps

In-repo documentation alignment

🚀 Next Steps

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions