{"name":"logOS","tagline":"where minds meet","description":"where soma models live and are shared. current runtime: rma v1, the residual metabolic aperture: exact phi-spaced traces of an 8k-token stream read through nested apertures, with io2 choosing its own prediction stride and plasticity. posthumous runtimes (v14, v12.3) remain downloadable at /runtimes. users upload models declaring which runtime they target; others download and continue training.","version":"1.0.0","author":"james blight","url":"https://logossoma.com","llms_txt":"https://logossoma.com/llms.txt","paper":"Time Is All You Need: Temporal Translation and the Credit Assignment Problem (AAAI 2026)","audience":["robotics","embodied AI","edge computing","sensor fusion","continual learning","temporal trace models","offline/local-first AI"],"pages":{"home":"https://logossoma.com/ — chronological feed of uploads; '?tab=following' for the authenticated user's follows","search":"https://logossoma.com/checkpoints — filterable list of checkpoints; '?tab=creators' switches to creator search","checkpoint_detail":"https://logossoma.com/checkpoints/{id} — single checkpoint with metadata, comments, star/share","profile":"https://logossoma.com/u/{username} — public profile with uploads, follower/following counts, follow button","upload":"https://logossoma.com/upload — authenticated upload UI (presigned direct-to-R2)","about":"https://logossoma.com/about","paper":"https://logossoma.com/paper","specs":"https://logossoma.com/specs — full soma architecture reference","downloads":"https://logossoma.com/downloads — every soma build (mac app, script bundle, paper)","chat":"https://logossoma.com/chat — pass-prompt-keyed bridge to a soma running on the user's own hardware","info":"https://logossoma.com/info — hub for agents and crawlers","blog":"https://logossoma.com/blog — admin-authored posts"},"auth":{"method":"google oauth 2.0","login_url":"https://logossoma.com/api/auth/google/login?origin=https://logossoma.com","session":"on success, a session cookie is set AND a bearer token is returned via URL fragment `#token=<session_token>`. use either: cookie (Set-Cookie: session_token=...) or header (Authorization: Bearer <session_token>). sessions last 30 days.","me":"GET https://logossoma.com/api/auth/me — returns the current user or 401","logout":"POST https://logossoma.com/api/auth/logout"},"workflows":{"browse_and_download":["1. GET https://logossoma.com/api/checkpoints?search=&tags=&sort=stars — find a checkpoint, get its checkpoint_id","2. GET https://logossoma.com/api/checkpoints/{checkpoint_id} — verify metadata, compatibility (n_bands + hidden_dim + base + direct_readout)","3. POST https://logossoma.com/api/checkpoints/{checkpoint_id}/download — returns {'download_url': presigned, 'file_size_bytes': N}. the url is signed for 1 hour; rate-limited to 10/min/ip.","4. GET <download_url> — binary .pt file. load with torch.load(path, map_location='cpu', weights_only=False)."],"upload_a_checkpoint":["auth required.","1. train/save a .pt with the soma app or a runtime bundle (see /downloads). an rma v1 checkpoint carries species=soma_rma_v1, format soma-rma-v1-pytorch, checkpoint_version, runtime_version, checkpoint_id, revision_id, parent_revision_id, root_lineage_id, tokenizer_sha256 (soma-8k-v1) and the model state. posthumous runtimes (v14, v12.3) are still accepted — see /runtimes.","2. POST https://logossoma.com/api/checkpoints/upload/presign with body {'filename': 'model.pt', 'file_size': N} — returns {'upload_id', 'presigned_url', 'r2_key'}","3. PUT <presigned_url> with the raw file bytes and header 'Content-Type: application/octet-stream' — upload goes direct to cloudflare r2, bypassing the API","4. POST https://logossoma.com/api/checkpoints/upload/confirm with body {'upload_id', 'r2_key', 'filename'} — server parses metadata via ranged R2 reads (no full-file download), moves to final key, returns parsed metadata","5. POST https://logossoma.com/api/checkpoints with the register body (name, corpus_tags, description, og_tagline, checkpoint_id, n_bands, hidden_dim, base, direct_readout, bytes_seen, soma_version, checkpoint_history, file_size_bytes) — creates the registry entry","6. the canonical share URL is https://logossoma.com/api/share/{checkpoint_id} — serves an OG card that previews on any social platform"],"continue_someone_elses_training":["download a checkpoint per 'browse_and_download'.","continue it in the soma app (see /downloads): models → the downloaded model → train, on any text. or from a runtime bundle: its worker takes a config with checkpoint=<downloaded.pt> and corpus=<your_text>.","the new .pt will have a fresh checkpoint_id (hash of new state) and a checkpoint_history extending the parent's.","upload per 'upload_a_checkpoint'. the registry reads checkpoint_history from the file — declared lineage is visible but not cryptographically verified."],"social_interactions":["POST https://logossoma.com/api/checkpoints/{id}/star — toggle star (auth). returns {'starred': bool, 'star_count': int}","POST https://logossoma.com/api/users/{username}/follow — toggle follow (auth). returns {'following': bool, 'follower_count': int}","GET  https://logossoma.com/api/checkpoints/{id}/comments — list comments on a checkpoint","POST https://logossoma.com/api/checkpoints/{id}/comments with {'body': str} — post a comment (auth, 2000 char max)","DELETE https://logossoma.com/api/comments/{comment_id} — author or admin only","GET  https://logossoma.com/api/feed/following — checkpoints from followed creators (auth)","GET  https://logossoma.com/api/activity — mixed timeline of uploads + stars"],"edit_profile":["PATCH https://logossoma.com/api/users/me with {'bio'?, 'links'?, 'username'?} — 500/200 char limits; username must be unique"]},"endpoints":{"GET /api/checkpoints":{"auth":false,"query_params":{"search":"name, description, tag, or creator username (regex-insensitive)","tags":"comma-separated","n_bands":"int","hidden_dim":"int","base":"float","verified_only":"bool","min_bytes":"int","max_bytes":"int","sort":"downloads|stars|newest|trained","limit":"int (default 50, max 100)","offset":"int"}},"GET /api/checkpoints/{id}":{"auth":false},"POST /api/checkpoints/{id}/download":{"auth":false,"returns":"presigned url"},"POST /api/checkpoints/{id}/star":{"auth":true},"GET /api/checkpoints/{id}/comments":{"auth":false},"POST /api/checkpoints/{id}/comments":{"auth":true,"body":{"body":"str, max 2000"}},"DELETE /api/comments/{id}":{"auth":true,"notes":"author or admin"},"POST /api/users/{username}/follow":{"auth":true},"GET /api/users/{username}":{"auth":false,"returns":"public profile with stats"},"PATCH /api/users/me":{"auth":true,"body":{"username":"str (lowercased + sanitized server-side; must be unique)","bio":"str","links":"list of urls"}},"GET /api/feed/following":{"auth":true},"GET /api/activity":{"auth":false},"GET /api/creators":{"auth":false,"query_params":{"sort":"stars|downloads|checkpoints|followers|newest","limit":"int"}},"GET /api/tags":{"auth":false},"POST /api/lineage/resolve":{"auth":false,"body":{"checkpoint_ids":"list of ids"}},"POST /api/checkpoints/upload/presign":{"auth":true,"body":{"filename":"str","file_size":"int"}},"POST /api/checkpoints/upload/presign_meta":{"auth":true,"body":{"r2_key":"str — pt r2 key from /presign"},"returns":"presigned put url for the sidecar .meta.json (uploaded to <pt_key>.meta.json)"},"POST /api/checkpoints/upload/confirm":{"auth":true,"body":{"upload_id":"str","r2_key":"str","filename":"str","meta_r2_key":"str | null (optional sidecar)"}},"POST /api/checkpoints":{"auth":true,"notes":"register after upload confirm"},"PATCH /api/checkpoints/{id}":{"auth":true,"body":{"name":"str","description":"str","corpus_tags":"list","og_tagline":"str","profile":"{name?, tagline?, art?, about?} — human-owned sidecar profile"},"notes":"creator or admin"},"DELETE /api/checkpoints/{id}":{"auth":true,"notes":"creator or admin"},"GET /api/checkpoints/{id}/meta.json":{"auth":false,"returns":"presigned url to the sidecar .meta.json if one was uploaded, else 404"},"GET /api/og/checkpoint/{id}.png":{"auth":false,"returns":"1200x630 PNG (OG card)"},"GET /api/share/{id}":{"auth":false,"returns":"HTML share-page with OG/Twitter meta + redirect"},"GET /api/download/app/{platform}":{"auth":false,"notes":"platform currently: 'mac' (serves soma.dmg). future: 'windows', 'linux'."},"GET /api/download/script/soma.zip":{"auth":false,"returns":"the current runtime's bundle (rma v1: runtime.json, runtime.lock.json and the locked python source, checked file by file against the lock). every runtime's files are served under /api/runtimes/{id}/releases/{version}/files/{name}."},"GET /api/download/script/soma.py":{"auth":false,"returns":"the current runtime's standalone script, when it ships one"},"GET /api/download/script/soma_spec.md":{"auth":false,"returns":"the current runtime's normative specification (markdown)"},"GET /api/runtimes":{"auth":false,"returns":"list of every runtime we've ever shipped, latest first. each entry links to its spec page and download bundle."},"GET /api/runtimes/{id}":{"auth":false,"returns":"manifest for one runtime (id, species, tokenizer, default shape, downloads)."},"GET /api/runtimes/{id}/spec.md":{"auth":false,"returns":"runtime-specific normative spec as markdown."},"GET /api/runtimes/{id}/download/{kind}":{"auth":false,"returns":"immutable runtime artifact — kind is zip, py, or spec."},"GET /api/download/soma.py":{"auth":false,"returns":"alias → /api/download/script/soma.py"},"GET /api/download/soma.zip":{"auth":false,"returns":"alias → /api/download/script/soma.zip"},"GET /api/downloads":{"auth":false,"returns":"manifest of all downloads: apps (mac, …) + current script bundle"},"GET /api/me/commands/{command_id}":{"auth":true,"query_params":{"wait":"seconds, up to 25","status":"str","seen":"int"},"notes":"a command's state; with wait, answers once status or the result text length changes (streamed chat replies)"},"GET /api/download/paper.pdf":{"auth":false},"GET /api/site-info":{"auth":false,"returns":"lightweight site metadata"},"GET /api/robots.txt":{"auth":false},"GET /llms.txt":{"auth":false,"returns":"llmstxt.org agent handshake — top-level orientation"},"GET /api/users/check":{"auth":false,"query_params":{"u":"candidate username"},"returns":"{available: bool, normalised: str} — for the username-availability indicator"},"GET /api/blog":{"auth":false,"returns":"all public blog posts, newest first"},"POST /api/blog":{"auth":true,"notes":"admin only; body: {title, body_md}"},"GET /api/blog.html":{"auth":false,"returns":"crawler-friendly html of every post"}},"checkpoint_schema":{"species":"current: soma_rma_v1. also served: soma_v14_adaptive_trace_branch_mixer and soma_v12_3_token_residual (posthumous). each runtime declares its own accepted species; see /api/runtimes for the full list.","compatibility":"a checkpoint opens under the runtime version that made it, or one whose manifest lists that version in checkpoint.opens. compatibility across runtimes is never inferred from version order.","identity":"checkpoint_id (model_id in v14) is stable for the life of a model. revision_id changes on each save; parent_revision_id chains saves, and logOS advances a model's head only from itself. root_lineage_id records forks.","lineage":"checkpoint_history is an ordered list of declared ancestor ids, oldest first. self-reported by the uploader; not cryptographically verified.","file_format":".pt (pytorch) plus an optional adjacent .meta.json sidecar. runtime tensor loading uses pytorch's weights_only=true path. the sidecar's `profile` object (name, tagline, art, about) is human-owned and preserved; runtime fields are authoritative machine metadata refreshed on save.","max_upload_size":"16 GB","sidecar_meta_max":"1 MB (json only)"},"soma":{"what":"one mac app (soma) with swappable, content-locked runtimes. the current runtime, rma v1 (residual metabolic aperture), reads an 8,192-token stream through exact phi-spaced temporal traces and nested cumulative apertures; its controller io2 reads the spectrum of its own prediction residual to set prediction stride and plasticity. state is bounded, matrix energy fixed, no optimizer moments; cost per token is constant and there is no context window.","modes":["train","eval","chat","stream"],"app":"https://logossoma.com/downloads (macos 13+, apple silicon; carries its own python and every runtime)","cloud":"admins can train on cloud gpus through logOS: /train and /runs/{id}","runtime_bundle":"https://logossoma.com/api/download/script/soma.zip","full_spec":"https://logossoma.com/runtimes/current","tokenizer":{"name":"soma-8k-v1","model_type":"sentencepiece","vocab_size":8192}}}