README: a decision tree for picking a setup - #140
Merged
Merged
Conversation
The README told you what the two modes measure, not which one you want. Adds a three-question flowchart (who is sending requests / how long are the prompts / do the answers quote the prompt back) with five lettered outcomes, and a table giving each one as a diff against the .env that Quick start already told you to copy — so "which setup" has a mechanical answer instead of a reading exercise. Writing it surfaced three things that were wrong: - .env.example suggested CTX=long while leaving SPEC=dflash2 set. DFlash2 past 64k is worth it only for context reproduction; SPEC=mtp CTX=long beats it about 2:1 on everything else (single-user/README.md). Says so now. - .env.example told CTX=huge users to run kvarn/install.sh first. That file is read only by Docker Compose, and the image already builds KVarN in. - single-user/README.md claimed CTX=huge "stays MTP (the script falls back with a message)". The launcher grew a dflash2+huge arm; docs/long-context.md has measured it at 268k tokens of pool since. Points there instead. Drops the "If you are the only user" section, whose content is now two rows of the table plus the link it already carried.
TyroneNel
pushed a commit
to TyroneNel/qwen38-27b-rtx3090
that referenced
this pull request
Sep 17, 2026
…i) into local main Upstream's six commits since the last sync: syv-ai#124, syv-ai#134, syv-ai#137, syv-ai#138 are the reviewed squashes of branches local main already carries (resolve_config.sh is byte-identical), so they land as the reviewed variants of the same features -- the launchers' fail-closed source guard, resolve_api_key.sh, and warmup.sh using that resolver instead of its own chain. syv-ai#140 and syv-ai#141 are new content: the README setup decision tree, .env.example, single-user/README.md and .github/FUNDING.yml. Conflicts, one hunk each: - batch/start_qwen.sh, single-user/start_qwen.sh: took upstream's guarded source (the launchers do not run under `set -e`); kept local's select_model.sh call in the single-user launcher. - docs/docker.md: kept local's paragraph (upstream never had it). - docs/gotchas.md: took upstream's indentation on gotcha 58's continuation. - docker/prepare.sh auto-merged into a duplicated translate block (local hoists DIRS and passes them to harden_chat_template.py; upstream nests them); kept local's version, a superset of upstream's. Verified: bash -n on every touched shell file, test_resolution.sh passes, patches/ untouched by the merge.
TyroneNel
added a commit
to TyroneNel/qwen38-27b-rtx3090
that referenced
this pull request
Sep 21, 2026
…i) into local main Upstream's six commits since the last sync: syv-ai#124, syv-ai#134, syv-ai#137, syv-ai#138 are the reviewed squashes of branches local main already carries (resolve_config.sh is byte-identical), so they land as the reviewed variants of the same features -- the launchers' fail-closed source guard, resolve_api_key.sh, and warmup.sh using that resolver instead of its own chain. syv-ai#140 and syv-ai#141 are new content: the README setup decision tree, .env.example, single-user/README.md and .github/FUNDING.yml. Conflicts, one hunk each: - batch/start_qwen.sh, single-user/start_qwen.sh: took upstream's guarded source (the launchers do not run under `set -e`); kept local's select_model.sh call in the single-user launcher. - docs/docker.md: kept local's paragraph (upstream never had it). - docs/gotchas.md: took upstream's indentation on gotcha 58's continuation. - docker/prepare.sh auto-merged into a duplicated translate block (local hoists DIRS and passes them to harden_chat_template.py; upstream nests them); kept local's version, a superset of upstream's. Verified: bash -n on every touched shell file, test_resolution.sh passes, patches/ untouched by the merge.
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.
The README explained what the two modes measure, but never answered the question a first-time reader actually has: which one do I want? That was spread across a comparison table, a prose section, and 70 lines of commented
.env.example.Adds a three-question flowchart — who is sending the requests, how long are the prompts, do the answers quote the prompt back — with five lettered outcomes, and a table giving each outcome as a diff against the
.envthat Quick start already told you to copy. Two of the five need no edit at all.Writing it exposed three stale or wrong claims, fixed here:
.env.examplesuggestedCTX=longwhile leavingSPEC=dflash2set above it. DFlash2 past 64k is worth it only for context reproduction —SPEC=mtp CTX=longbeats it about 2:1 on everything else, whichsingle-user/README.mdhas said for a while..env.exampletoldCTX=hugeusers to runbash kvarn/install.shfirst. That file is read only by Docker Compose, andDockerfile:38already builds KVarN into the image. Only the venv install needs the step.single-user/README.mdsaidCTX=huge"stays MTP (the script falls back with a message)".single-user/start_qwen.sh:204grew adflash2+hugearm, anddocs/long-context.mdhas measured it at 268k tokens of pool at 245760 max-model-len.The "If you are the only user" section is dropped: it is now two rows of the table plus the link it already carried. Net +22 lines on the README.
Checked: the flowchart renders (
mmdcproduces the SVG), and the repo-wide link/anchor check is still 0 broken.