From the repository, docs/README.md
Documentation
Each page covers one part of TandemLLM and names the files that implement it. Start with the architecture page, a one-page tour that links to all the others. To install, start with setup.md. The manual, tandemllm-manual.pdf, puts these pages and the reference tables in one book.
| page | what it covers |
|---|---|
| architecture.md | the 27B and Flash-Next, the byte budget, the decode loop, the layers every model plugs into, a map of the code |
| quantisation.md | the checkpoint's FP8, NVFP4, the clip search, GPTQ from the BF16 release, the quality gate, the three profiles, Flash-Next's EXL3 packs |
| kernels.md | row invariance, the 4-bit and 8-bit matrix products, the recurrent-layer kernels, graphs |
| speculative-decoding.md | the block drafter, the lookup drafter, StairCut, the older length router, tree verify through recurrent layers, Flash-Next's MTP drafter |
| exactness.md | the promise about the text, the near-tie exception, and the checks behind it |
| caches.md | the resident prefix, the state store, the persistent suffix store |
| server.md | the OpenAI API, tool calls, structured outputs, streaming, admission |
| parallel-requests.md | serving several requests at once (--parallel N): switching it on, memory, speed, the text, Flash-Next |
| dashboard.md | the Live activity view, the other views, the API contract |
| packages.md | the packages the conductor attaches by class: their shape, the first packages, how each is measured |
| operations.md | running a profile, memory rules, holds, deploys, troubleshooting |
| measurement.md | the benchmark row, the teacher-forced bench, noise, the release gate |
| benchmarks.md | tool-eval-bench against TandemLLM: the raw and conductor paths, the settings a score must name, results on a DGX Spark |
| tool-eval-bench.md | tool-eval-bench results of 2026-10-19: the raw path at five thinking settings, four conductor profiles, hard mode |
| adding-a-model.md | the adapter path a new model takes, the seams, and the tests it must pass |
| lfm2.5-24b-a2b.md | LiquidAI's LFM2.5-24B-A2B: the model, its profiles, its tool-call format, how the engine runs it |
| nemotron-h.md | NVIDIA's NemotronH (Nemotron 3.5 Bolt, Super VL, Lightning): the models, the EXL3, GGUF and NVFP4 profiles, LiLiCorr, how the engine runs them |
| setup.md | a new box from the command line, the dashboard's setup guide, the model registry behind it, and Wren |
| chat.md | the Chat tab (in beta) and chatd, the service behind it: the sandbox, the permission modes, the known limitations |
| community-weights.md | packs other people published: running one, adding yours with one JSON file, how a maintainer verifies it |
| memory.md | memory notes: what the model keeps about the user and the work |
| shadow-reader.md | the small model that follows each session and answers questions about it |
| proxy.md | the dashboard as a proxy in front of other engines and cloud APIs |
| roadmap.md | what Flash-Next still lacks, parallel requests, other GPUs |
Two more references live next to the code. server/METRICS.md lists every Prometheus metric, and contract/dashboard-v1/ holds the dashboard API's schemas.
Every number on these pages was measured on one DGX Spark, unless the page calls it arithmetic or an estimate. measurement.md says how.
This documentation is licensed under CC BY 4.0 (LICENSE). The code is licensed separately (see the README).