docs(adr): ESP32-S3 micro-LLM research spike + flash-resident weight streaming (ADR-322/328) - #1680
Open
Esbroro wants to merge 2 commits into
Open
docs(adr): ESP32-S3 micro-LLM research spike + flash-resident weight streaming (ADR-322/328)#1680Esbroro wants to merge 2 commits into
Esbroro wants to merge 2 commits into
Conversation
Adds a source-cited research survey of the 2026 ESP32-S3 micro-LLM demonstrations (slvDev esp32-ai, DaveBben esp32-llm, llama2.c, Gemma-3n Per-Layer Embeddings) with all external numbers tagged CLAIMED, plus: - ADR-324: micro-LLM inference is a bounded research spike on dedicated 16MB-flash companion hardware and an explicit production non-goal for CSI node firmware (flash-impossible on the 8MB/4MB fleet, core/PSRAM contention with sensing, no task capability at TinyStories scale). - ADR-325: design investigation into flash-resident memory-mapped quantized weight streaming for RuView's own edge models, with a measured-benchmark-first gate and provenance requirements. No firmware, partition, or wire changes; documentation only. Co-Authored-By: claude-flow <ruv@ruv.net> Claude-Session: https://claude.ai/code/session_01KCYFXcPDG6yTCxwefb6iy6
The research spike (ADR-324) and flash-resident weight streaming (ADR-325) ADRs were authored on 2026-08-16 on a branch that fell behind main. By the time it's picked back up, ADR-324 through ADR-327 were assigned to unrelated work (off-axis perspective demo, Cognitum Spaces). Renumber to the next free slots: ADR-322 (an observed gap) and ADR-328 (next after ADR-327), fix internal cross-references, and index both in docs/adr/README.md. Both ADRs remain Status: Proposed, unreviewed content unchanged. Co-Authored-By: claude-flow <ruv@ruv.net>
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
2 participants
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.
Summary
Recovers and lands a bounded research spike into on-device micro-LLM inference on ESP32-S3, originally drafted 2026-08-16 on an unmerged branch that fell behind
main.firmware/esp32-csi-node/under this ADR. All external performance numbers stayCLAIMEDuntil reproduced on RuView silicon with a witness log.MEASUREDADR.docs/research/esp32-micro-llm-inference.md: companion research doc with sources, technique analysis, and fleet fit-analysis backing both ADRs.Renumbering
The branch was authored as ADR-324/ADR-325, but by the time it was picked back up
mainhad assigned ADR-324 through ADR-327 to unrelated work (off-axis perspective demo, Cognitum Spaces). Renumbered to the next free slots — ADR-322 (an observed gap in the index) and ADR-328 (next free after ADR-327) — fixed all internal cross-references, and indexed both indocs/adr/README.md. Decision content is unchanged from the original commit; both remainStatus: Proposed.Non-goals reiterated
🤖 Generated with claude-flow