🔴 Live TV: lemonade sdk/lemonade/pull/

sawansri · 2026-03-05T20:54:37Z

Implements lemonade-server launch for easy integration of lemonade with coding agents like Claude Code and Codex.

Review checklist

Clarity

Does the pr description and title explain the change fully?
Are UI/UX changes displayed with screenshots/videos/terminal blocks?

Completeness

Are changes documented in markdown and help menus (if appropriate)?
- For example: cli.md, spec.md, dev-getting-started.md, lemonade-server --help
Do new endpoints have example requests/scripts so others can reproduce?

UI/UX review

Are the UI/UX changes (if any)...

visually cohesive with the rest of the experience?
a cohesive part of existing workflows, or do they introduce a new workflow (justify)?
discoverable by new users?
usable without documentation, or is new documentation needed (justify)?
tested on both Windows on Linux (if needed)?

Note: UI/UX includes installers, app, tray, and CLI.

Code quality

Is the change maintainable (does it introduce redundancy)?
Are there any security or licensing issues?

superm1

General comment - I am wondering if we actually want a compatibility symlink for /usr/bin/lemonade and /usr/bin/lemonade-server and then in docs tell people to use lemonade instead.

sawansri · 2026-03-05T21:08:25Z

General comment - I am wondering if we actually want a compatibility symlink for /usr/bin/lemonade and /usr/bin/lemonade-server and then in docs tell people to use lemonade instead.

I think that makes sense. In my mind, launch is less of a "server" feature and fits better with the "lemonade" binary.

sawansri · 2026-03-05T22:13:29Z

Hey guys, something I want to get thoughts on is when the model should be loaded in llama-server. The current implementation loads the model asynchronously while the agent is launched (versus an earlier implementation when it was blocking agent startup). Is this a good approach?

superm1 · 2026-03-05T22:15:52Z

I guess what happens if the model load isn't done by the time the agent starts up? Race? Chaos? If so synchronous is the right way.

sawansri · 2026-03-05T22:26:48Z

I guess what happens if the model load isn't done by the time the agent starts up? Race? Chaos? If so synchronous is the right way.

Based on the testing I did, seems like the processing of the query is just delayed until the model is finished loading. Imo, async leads to a smoother experience since the user can potentially start typing out their prompt while the model is simultaneously loading in the backend.

superm1 · 2026-03-05T22:36:11Z

I guess what happens if the model load isn't done by the time the agent starts up? Race? Chaos? If so synchronous is the right way.

Based on the testing I did, seems like the processing of the query is just delayed until the model is finished loading. Imo, async leads to a smoother experience since the user can potentially start typing out their prompt while the model is simultaneously loading in the backend.

Yeah I tend to agree with you from that description. People expect the first query is slow anyways.

sawansri · 2026-03-06T04:27:53Z

Failing test seems to be related to a missing windows import (processenv.h which is properly located in the header guards).

Signed-off-by: Sawan Srivastava <sawan1210@gmail.com>

… does Signed-off-by: Sawan Srivastava <sawan1210@gmail.com>

jeremyfowers · 2026-03-06T22:05:47Z

General comment - I am wondering if we actually want a compatibility symlink for /usr/bin/lemonade and /usr/bin/lemonade-server and then in docs tell people to use lemonade instead.

@bitgamma is looking into this!

superm1 · 2026-03-06T23:14:58Z

@jeremyfowers can you help figure out the Windows issue? If we can I'd like to see this in 10.0.0!

…uments, and --use-recipe flag to prioritize model recipe Signed-off-by: Sawan Srivastava <sawan1210@gmail.com>

Signed-off-by: Sawan Srivastava <sawan1210@gmail.com>

The lemonade-server target defines UNICODE/_UNICODE, which causes GetEnvironmentStringsA to be undefined (it's only a macro when UNICODE is not set). Switch to explicit wide-string versions (GetEnvironmentStringsW/ FreeEnvironmentStringsW) with UTF-8 conversion via WideCharToMultiByte. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

jeremyfowers · 2026-03-09T20:15:41Z

@claude review

claude · 2026-03-09T20:15:56Z

Claude Code is working…

I'll analyze this and get back to you.

View job run

jeremyfowers

Works great for me! Agreed to two followups. Thanks @sawansri!

* basic launch implementation Signed-off-by: Sawan Srivastava <sawan1210@gmail.com> * added asynchronous loading of model when agent is launched Signed-off-by: Sawan Srivastava <sawan1210@gmail.com> * switch to LOG and use helper for host normalization Signed-off-by: Sawan Srivastava <sawan1210@gmail.com> * add some docs for launch subcommand Signed-off-by: Sawan Srivastava <sawan1210@gmail.com> * missing header file? Signed-off-by: Sawan Srivastava <sawan1210@gmail.com> * improve prompt caching by disabling system prompt changes claude code does Signed-off-by: Sawan Srivastava <sawan1210@gmail.com> * implements sensible llamacpp defaults, user configurable llamacpp arguments, and --use-recipe flag to prioritize model recipe Signed-off-by: Sawan Srivastava <sawan1210@gmail.com> * disable telemetry and other non-essential traffic Signed-off-by: Sawan Srivastava <sawan1210@gmail.com> * Fix UNICODE build error in process_manager for lemonade-server The lemonade-server target defines UNICODE/_UNICODE, which causes GetEnvironmentStringsA to be undefined (it's only a macro when UNICODE is not set). Switch to explicit wide-string versions (GetEnvironmentStringsW/ FreeEnvironmentStringsW) with UTF-8 conversion via WideCharToMultiByte. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Signed-off-by: Sawan Srivastava <sawan1210@gmail.com> Co-authored-by: Jeremy Fowers <80718789+jeremyfowers@users.noreply.github.com> Co-authored-by: Jeremy Fowers <jeremy.fowers@amd.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

superm1 reviewed Mar 5, 2026

View reviewed changes

Comment thread src/cpp/tray/tray_app.cpp Resolved

superm1 reviewed Mar 5, 2026

View reviewed changes

Comment thread src/cpp/tray/tray_app.cpp Resolved

superm1 reviewed Mar 5, 2026

View reviewed changes

Comment thread src/cpp/tray/tray_app.cpp Resolved

superm1 reviewed Mar 5, 2026

View reviewed changes

Comment thread src/cpp/server/cli_parser.cpp Resolved

superm1 approved these changes Mar 5, 2026

View reviewed changes

sawansri marked this pull request as ready for review March 5, 2026 22:29

sawansri added 6 commits March 6, 2026 12:46

basic launch implementation

162d886

Signed-off-by: Sawan Srivastava <sawan1210@gmail.com>

added asynchronous loading of model when agent is launched

50be49e

Signed-off-by: Sawan Srivastava <sawan1210@gmail.com>

switch to LOG and use helper for host normalization

b683885

Signed-off-by: Sawan Srivastava <sawan1210@gmail.com>

add some docs for launch subcommand

3b4600a

Signed-off-by: Sawan Srivastava <sawan1210@gmail.com>

missing header file?

ea05c81

Signed-off-by: Sawan Srivastava <sawan1210@gmail.com>

improve prompt caching by disabling system prompt changes claude code…

938a42e

… does Signed-off-by: Sawan Srivastava <sawan1210@gmail.com>

sawansri force-pushed the launch branch from 6fd20e1 to 938a42e Compare March 6, 2026 20:46

sawansri added 2 commits March 6, 2026 15:59

implements sensible llamacpp defaults, user configurable llamacpp arg…

e24613c

…uments, and --use-recipe flag to prioritize model recipe Signed-off-by: Sawan Srivastava <sawan1210@gmail.com>

disable telemetry and other non-essential traffic

7a96c58

Signed-off-by: Sawan Srivastava <sawan1210@gmail.com>

sawansri force-pushed the launch branch from b3c59ab to 7a96c58 Compare March 7, 2026 03:47

jeremyfowers assigned sawansri Mar 9, 2026

Merge branch 'main' into launch

78a4214

jeremyfowers reviewed Mar 9, 2026

View reviewed changes

Comment thread docs/server/lemonade-server-cli.md Resolved

jeremyfowers mentioned this pull request Mar 9, 2026

Launch command followup: UI/UX #1331

Closed

jeremyfowers approved these changes Mar 9, 2026

View reviewed changes

ramkrishna2910 mentioned this pull request Mar 9, 2026

Fix claude-code-action failing on fork PRs #1332

Merged

2 tasks

jeremyfowers merged commit 3cf6e25 into lemonade-sdk:main Mar 9, 2026
38 checks passed

sawansri mentioned this pull request Mar 10, 2026

Add a user guide for Claude Code integration #1334

Merged

Conversation

sawansri commented Mar 5, 2026 • edited by jeremyfowers Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review checklist

Clarity

Completeness

UI/UX review

Code quality

Uh oh!

superm1 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

sawansri commented Mar 5, 2026

Uh oh!

sawansri commented Mar 5, 2026

Uh oh!

superm1 commented Mar 5, 2026

Uh oh!

sawansri commented Mar 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

superm1 commented Mar 5, 2026

Uh oh!

sawansri commented Mar 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jeremyfowers commented Mar 6, 2026

Uh oh!

superm1 commented Mar 6, 2026

Uh oh!

Uh oh!

jeremyfowers commented Mar 9, 2026

Uh oh!

claude Bot commented Mar 9, 2026

Uh oh!

jeremyfowers left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

sawansri commented Mar 5, 2026 •

edited by jeremyfowers

Loading

sawansri commented Mar 5, 2026 •

edited

Loading

sawansri commented Mar 6, 2026 •

edited

Loading