Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
67 commits
Select commit Hold shift + click to select a range
e12ed39
feat(examine,serve): report real GPU/backend support instead of a sta…
r0x0r Oct 14, 2026
c63fa43
test(serve): keep the api key out of summary assertion messages
r0x0r Oct 14, 2026
67cd482
test(runtime): assert the Vulkan-serving remediation where the refusa…
r0x0r Oct 14, 2026
61bb3c3
fix(install): stop the requested-target refusal predicting a backend
r0x0r Oct 14, 2026
9f78b66
fix(bridge): report lemonade's real backend in the codex-bridge inven…
r0x0r Oct 16, 2026
8ada511
Merge branch 'main' into ROCMAI-550-vulkan-support-messaging
r0x0r Oct 16, 2026
18666d4
test(engines): cover the GPU-aware engine inventory on `rocm engines`…
r0x0r Oct 16, 2026
1b85b9e
Merge branch 'main' into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 2, 2026
a9ed8fa
ROCMAI-550: make the device row name a device and examine answer ever…
r0x0r Nov 2, 2026
7fba9aa
ROCMAI-550: resync the advice-scan exclusions with the new device shape
r0x0r Nov 2, 2026
e1daa50
ROCMAI-550: drop the engine advice-scan exclusions the shared formatt…
r0x0r Nov 2, 2026
a3f27af
Merge origin/main into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 2, 2026
6d79b3d
Merge origin/main (#703) into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 3, 2026
f4e02f7
ROCMAI-550: pin the support line with a scenario, and stop the device…
r0x0r Nov 3, 2026
11819e9
Merge origin/main into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 3, 2026
68bea03
Merge origin/main into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 3, 2026
f72bfb3
Merge origin/main into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 4, 2026
ea43070
Merge origin/main into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 5, 2026
73ec5ae
Merge origin/main, and restore the serve --help sentence the last mer…
r0x0r Nov 5, 2026
c65224a
fix(examine): point the unknown-tier advice at a surface that shows t…
r0x0r Nov 6, 2026
34845bf
fix(examine): stop sending a GPU-less host to a command serve would r…
r0x0r Nov 9, 2026
c806ee3
Merge origin/main, and follow this PR's changes to where main moved them
r0x0r Nov 9, 2026
8ffeca6
test(serve): show the Lemonade advice names a command this host is no…
r0x0r Nov 10, 2026
5de5b43
Merge remote-tracking branch 'origin/main' into pr678-r2
r0x0r Nov 10, 2026
b653cc8
test(engine): assert the endpoint key file is withheld from plugin ch…
r0x0r Nov 10, 2026
a9d03d1
Merge remote-tracking branch 'origin/main' into pr678-r2
r0x0r Nov 10, 2026
130747e
Merge origin/main into pr678-r2
r0x0r Nov 11, 2026
6a92843
feat(examine): probe Vulkan instead of inferring it from the gfx table
r0x0r Nov 11, 2026
129b043
fix(engine): clear the endpoint key file for a child that gets no key
r0x0r Nov 11, 2026
b6c37c6
docs(services): condense the device-probe doc blocks to clear the siz…
r0x0r Nov 11, 2026
61ae34a
docs(examine): correct the support-level section and move items above…
r0x0r Nov 11, 2026
d3b3e34
Merge remote-tracking branch 'origin/main' into pr678-r2
r0x0r Nov 11, 2026
654c2de
fix(engine,examine): scope the key-file clearing, and stop claiming V…
r0x0r Nov 11, 2026
3754e92
refactor(examine): split the Vulkan probe and the protocol change out…
r0x0r Nov 11, 2026
67c27ea
Merge remote-tracking branch 'origin/main' into pr678-r2
r0x0r Nov 11, 2026
c03972e
fix(examine): finish removing the split-out work from this PR
r0x0r Nov 11, 2026
545da1f
docs(examine): say the support line is printed on every host
r0x0r Nov 11, 2026
db00642
fix(cli): keep the new examine help text ASCII-only
r0x0r Nov 11, 2026
2ce0e14
fix(examine): stop naming a backend for a GPU the catalog does not know
r0x0r Nov 11, 2026
9b5788e
Merge remote-tracking branch 'origin/main' into pr678-r2
r0x0r Nov 11, 2026
3631908
test(examine): split the examine-49 arm the Unknown-tier advice outgrew
r0x0r Nov 12, 2026
c618148
docs(testing): describe the levels this change has, not an unmerged one
r0x0r Nov 12, 2026
c2446ba
Merge origin/main into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 12, 2026
f050eb1
fix(serve): skip the device probe for a service that was already running
r0x0r Nov 13, 2026
e6c0df9
Merge origin/main into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 13, 2026
64dc074
fix(engines): stop `rocm engines` requiring ROCm on a host it says ha…
r0x0r Nov 13, 2026
cc9e90a
fix(cli): stop markdown markup reaching `--help`, and make the Vulkan…
r0x0r Nov 13, 2026
39bec51
docs: point the static-catalog note at rocmd's real engine inventory
r0x0r Nov 13, 2026
de1e3f3
Merge branch 'main' into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 16, 2026
81a5f5b
Merge origin/main into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 16, 2026
557aa34
fix(services): stop the device probe losing a just-published anthropi…
r0x0r Nov 16, 2026
4608e96
docs(services): scope the device-probe wait to the table, not `--json`
r0x0r Nov 16, 2026
074becf
fix(examine): stop an SDK for another family being called this GPU's
r0x0r Nov 16, 2026
d99fd16
Merge origin/main into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 16, 2026
414bf50
Merge origin/main into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 17, 2026
4a651b5
fix(serve): stop the engine's post-spawn state rewrite dropping the A…
r0x0r Nov 17, 2026
aca4f36
Merge origin/main into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 17, 2026
cd895c7
Merge origin/main into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 17, 2026
132394e
Merge origin/main into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 17, 2026
017a109
fix(engines): stop AIM claiming a GPU it was never given, and write s…
r0x0r Nov 17, 2026
4e435f7
test(core): reach the two branches the GPU classifier's own suite cou…
r0x0r Nov 17, 2026
b8a064b
test(e2e): say that the cross-surface device check cannot fire, and l…
r0x0r Nov 17, 2026
5ba7eb2
docs: correct the device-row gating, the gfx_target entry and the Lem…
r0x0r Nov 17, 2026
9e0f3c1
Merge origin/main into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 17, 2026
f04357f
Merge origin/main into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 17, 2026
7bfee25
Merge origin/main into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 17, 2026
06ece66
Merge origin/main into ROCMAI-550-vulkan-support-messaging
r0x0r Nov 17, 2026
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
46 changes: 38 additions & 8 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -682,6 +682,12 @@ family is substituted for it. `rocm examine` reports the same conclusion on a

These GPUs still work with the Vulkan backend, which needs no ROCm SDK.

`rocm examine` prints that `rocm_support` line on **every** host, not only this
case, at one of three levels: ROCm/HIP-supported, Vulkan-only (the case above),
or unknown. Unknown covers both no GPU detected and a target this build's
catalog does not recognise — detection gaps rather than a verdict about the
hardware, so neither is reported as "no ROCm build".

#### Compiler toolchain (--devel)

Pass `--devel` to also install the compiler and headers needed to build GPU
Expand Down Expand Up @@ -1390,18 +1396,33 @@ the running service, or stop it first to serve on the new port.

By default the server runs in the background under ROCm CLI's supervision and
prints a deployment summary: a progress indicator while it starts, then a table
with the status, the full inference endpoint, the API-qualified model name, and a
quick smoke test (time to first token and approximate tokens per second). On an
interactive terminal, if the smoke test cannot complete, the summary is headed
`(smoke test failed)`, shows the status as `started, smoke test failed` instead
of `ready`, and a note gives the reason and points at
with the status, the engine and the backend device the engine reports (for
example `rocm gpu 0 (AMD Instinct MI300X)`, or `vulkan (AMD Radeon Graphics)`
where the engine pinned no GPU ordinal). It is read back from the running
engine, so it is that engine's own account of the backend it was launched on
rather than an independent measurement of the hardware. The summary also carries
the full inference endpoint, the API-qualified model name, and a quick smoke
test (time to first token and approximate tokens per second). `rocm services`
reports the same `device` for each running service.

On an interactive terminal, if the smoke test cannot complete, the summary is
headed `(smoke test failed)`, shows the status as `started, smoke test failed`
instead of `ready`, and a note gives the reason and points at
`rocm logs --service <id>`. ROCm CLI does not stop the server, and the command
still exits 0. Which failures are detected, and which still read as `ready`, is
listed under
[Serve Smoke Test Failure](https://github.com/ROCm/rocm-cli/blob/main/docs/testing.md#serve-smoke-test-failure)
in the testing guide. Control
returns to your shell with the server still running; manage it later with `rocm
services` (below).
in the testing guide.

The device line is omitted when the engine cannot say yet. A server is reported
ready as soon as it answers, which can be before it has finished loading the
model, and until then the engine does not know which device it ended up on;
`rocm serve` waits a few seconds for it rather than holding your terminal, and

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

"Waits a few seconds" understates it: the code is a 6s start gate plus one in-flight probe, about 14s per the engine_ipc.rs comment, on a launch that ends without a row. Please align this with docs/testing.md and the code.

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

"Waits a few seconds" understates it: the code is a 6s start gate plus one in-flight probe (about 14s per the engine_ipc.rs comment) on a launch that may then print no row. Align this with docs/testing.md and the code.

`rocm services` shows it once it is known. The GPU name is omitted on a machine
with more than one AMD GPU, where naming one of them could name the wrong one.

Control returns to your shell with the server still running; manage it later
with `rocm services` (below).

`--verbose` (or `--foreground`) instead attaches to the server in the current
terminal and streams every engine log line. Use it to debug a startup problem.
Expand Down Expand Up @@ -1646,6 +1667,15 @@ It does not suggest running the stop again: when only a process under the
recorded one survived, a second stop can report the server stopped while that
process still runs.

`list` asks each running server which backend and GPU it is actually on, so it
can take a few seconds rather than returning instantly. The query is capped in

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

"Capped in total ... however many servers" isn't true for external plugin engines: the code doc on fetch_service_devices says engine_request waits on a plugin child unbounded. Please say "built-in engines" here and in the services list --help.

total — about 18 seconds worst case, however many servers are listed — and a

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

"About 18 seconds worst case, however many servers are listed" only holds for built-in engines. engine_ipc.rs waits on an external plugin child with no deadline, so rocm services list (and the chat services tool) can still hang on a ready plugin-backed service. Either narrow this wording (and the ServicesCommand::List help in cli.rs) to built-in engines, or skip the device probe for non-built-in engines.

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

"About 18 seconds worst case, however many servers are listed" only holds for built-in engines: engine_ipc.rs documents that engine_request waits on an external plugin child with no deadline, so a ready plugin-backed service can still hang rocm services list and the chat services command. Narrow this wording (and the ServicesCommand::List help in cli.rs) to built-in engines, or skip the probe for non-built-in engines.

server that does not answer in time simply has no `device` line rather than
holding up the rest of the list. The same applies to asking about servers in
`rocm chat`, which prints the same list. `list --json` is unaffected: it emits
the stored service records, which have no `device` field, asks no server
anything, and returns as fast as it always did.

`restart` starts a server again only once its stop has confirmed that every
recorded process is gone; if the stop cannot confirm that, it says so, starts
nothing, and leaves the stop request recorded.
Expand Down
18 changes: 18 additions & 0 deletions apps/rocm/src/advised_commands.rs
Original file line number Diff line number Diff line change
Expand Up @@ -1304,6 +1304,24 @@ const NOT_INVOCATIONS: &[(&str, &str, &str)] = &[
"the doc's expected result is that the parser rejects the unknown option \
with exit 2, the code the catalog publishes for an unknown fix-id",
),
// The `device:` row's value (ROCMAI-550). Every engine reports the backend
// it selected as `<backend> [gpu <n>] [(<name>)]` -- `vulkan (AMD Radeon
// Graphics)`, `rocm gpu 1 (AMD Instinct MI300X)` -- so the ROCm half of
// that value reads as a `rocm` invocation to this scan while naming no
// command at all. The engines no longer need an entry each: they build the
// string through `rocm_engine_protocol::format_reported_device` rather
// than holding a `"rocm gpu ..."` literal. These are the two docs that
// quote the format back.
(
"README.md",
"rocm gpu 0 (AMD Instinct MI300X)",
"an example `device:` value in the deployment-summary description, not a command",
),
(
"docs/manual-testing.md",
"rocm gpu 0 (AMD Instinct MI300X)",
"the `device:` value the tester is told to expect, not a command to run",
),
];

/// Natural-language examples must reach the planner with a multi-word request:
Expand Down
32 changes: 28 additions & 4 deletions apps/rocm/src/bridge_snapshot.rs
Original file line number Diff line number Diff line change
Expand Up @@ -32,15 +32,24 @@ use crate::managed_service_status::load_managed_services;

pub(crate) fn build_codex_bridge_snapshot(paths: &AppPaths) -> Result<CodexBridgeSnapshot> {
let config = RocmCliConfig::load(paths).unwrap_or_default();
// Hoisted out of the struct literal so the engine rows can be told what this
// host actually supports. The probe is already paid for: `gpu_support_level`
// takes exactly the two fields this summary already carries, so the engine
// list costs no detection the snapshot was not doing anyway.
let examine = ExamineSummary::gather()?;
let gpu_support_level = rocm_core::gpu_support_level(

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This positional gpu_support_level(detected, family) call, and the one at examine_report_text.rs:136, bypass gpu_support_level_for_host_gpu, whose doc exists because the two same-typed args can be transposed. A single ExamineSummary::gpu_support_level() would keep both call sites honest.

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This positional gpu_support_level(detected, family) call, and the one in examine_report_text.rs, bypass gpu_support_level_for_host_gpu, whose doc exists because the two same-typed Option<&str> args can be transposed without a compile error. One ExamineSummary::gpu_support_level() would keep every caller honest.

examine.detected_gfx_target.as_deref(),
examine.compatible_therock_family.as_deref(),
);

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Blocking (test coverage) — this gpu_support_level wiring is new behavior this PR introduces, and nothing tests it.

build_codex_bridge_snapshot computes gpu_support_level from ExamineSummary::gather()'s two fields, independently of main.rs's host_gpu_support_level (which uses detect_host_gpu_summary(paths)). Verified: this file has zero #[test] functions. The sibling renderer builtin_codex_bridge_engine_inventory (used downstream) is well-tested via chat/mcp_bridge.rs's internal_mcp_engines_tool_result, but that test injects a pre-computed GpuSupportLevel directly, bypassing this exact call site — so a positional transposition of the two Option<&str> arguments at lines 40-43 (a bug class gpu_support_level's own doc comment explicitly warns about) would compile and pass every existing test.

Confidence 72. Suggested fix: add a test that plants an ExamineSummary (or exercises build_codex_bridge_snapshot end-to-end) and asserts the resulting snapshot's engine notes match the expected GpuSupportLevel.

Ok(CodexBridgeSnapshot {
protocol: "rocmd-codex-bridge-v0".to_owned(),
generated_at_unix_ms: rocm_core::unix_time_millis(),
examine: ExamineSummary::gather()?,
examine,
gpu: build_codex_bridge_gpu_snapshot(&config),
config,
automation_runtime: AutomationRuntimeState::load(paths)?,
recent_automation_events: load_recent_automation_events(paths, 32)?,
engines: builtin_codex_bridge_engine_inventory(),
engines: builtin_codex_bridge_engine_inventory(gpu_support_level),
services: load_managed_services(paths)?,
})
}
Expand All @@ -63,7 +72,22 @@ pub(crate) fn build_codex_bridge_gpu_snapshot(config: &RocmCliConfig) -> CodexBr
}
}

pub(crate) fn builtin_codex_bridge_engine_inventory() -> Vec<CodexBridgeEngine> {
/// The engine catalogue in the codex-bridge snapshot and the bridge's `engines`
/// tool, built from [`engine_inventory`] with lemonade's note replaced
/// by `lemonade_engine_note` for the given support level.
///
/// It takes the level rather than reading the static row because the stale
/// "with ROCm llama.cpp backend" claim is exactly what ROCMAI-550 exists to
/// remove, and this is a user-facing surface like the other two: a bridge
/// client renders `summary` to the operator.
///
/// Only `summary` moves. The field set, its names and `installed_binary: true`
/// beside it are untouched, so the bridge's JSON shape is the same; what
/// changes is that one string stops asserting ROCm support on a host that has
/// none.
pub(crate) fn builtin_codex_bridge_engine_inventory(
gpu_support_level: rocm_core::GpuSupportLevel,
) -> Vec<CodexBridgeEngine> {
let default_engine = default_engine_for_platform();
let binary_path = daemon_binary_path()
.ok()
Expand All @@ -72,7 +96,7 @@ pub(crate) fn builtin_codex_bridge_engine_inventory() -> Vec<CodexBridgeEngine>
.iter()
.map(|(id, summary)| CodexBridgeEngine {
id: (*id).to_owned(),
summary: (*summary).to_owned(),
summary: crate::engine_inventory_note(id, summary, gpu_support_level).to_owned(),
default_for_platform: *id == default_engine,
installed_binary: true,
binary_path: binary_path.clone(),
Expand Down
102 changes: 92 additions & 10 deletions apps/rocm/src/chat/mcp_bridge.rs
Original file line number Diff line number Diff line change
Expand Up @@ -44,9 +44,10 @@ use crate::uninstall_plan::render_uninstall_dry_run;
use crate::update_cmd::render_update_text;
use crate::{
build_codex_bridge_gpu_snapshot, build_codex_bridge_snapshot,
builtin_codex_bridge_engine_inventory, cli_version_string, record_cli_audit_event,
render_config_text, render_engine_inventory_text_with_paths, render_examine_text,
render_examine_text_with_paths, render_model_registry_text_with_context_and_host,
builtin_codex_bridge_engine_inventory, cli_version_string, host_gpu_support_level,
record_cli_audit_event, render_config_text, render_engine_inventory_text_with_paths,
render_examine_text, render_examine_text_with_paths,
render_model_registry_text_with_context_and_host,
render_model_registry_verbose_text_with_context_and_host, render_setup_status_text,
therock_install_version_selector,
};
Expand Down Expand Up @@ -186,13 +187,9 @@ pub(crate) fn run_internal_mcp_call(
serde_json::json!(gpu),
))
}
"engines" => {
let engines = builtin_codex_bridge_engine_inventory();
Ok(internal_mcp_tool_success(
format!("Found {} engine entries.", engines.len()),
serde_json::json!({ "engines": engines }),
))
}
"engines" => Ok(internal_mcp_engines_tool_result(host_gpu_support_level(
Some(paths),
))),
"services" => {
let services = load_managed_services(paths)?;
Ok(internal_mcp_tool_success(
Expand Down Expand Up @@ -388,6 +385,24 @@ pub(crate) fn run_internal_mcp_call(
}
}

/// The `engines` MCP tool result for a host at `gpu_support_level`.
///
/// Split from the match arm above only so the GPU-aware branch is reachable
/// from a test: the arm's own argument is a live host probe, and no CI runner
/// has a GPU that is detected *and* confirmed to have no ROCm build, so the
/// Vulkan-only wording a bridge client would render could not otherwise be
/// asserted on. Nothing else moved — the summary text, the `engines` key and
/// the entries themselves are what the arm built inline.
fn internal_mcp_engines_tool_result(
gpu_support_level: rocm_core::GpuSupportLevel,
) -> serde_json::Value {
let engines = builtin_codex_bridge_engine_inventory(gpu_support_level);
internal_mcp_tool_success(
format!("Found {} engine entries.", engines.len()),
serde_json::json!({ "engines": engines }),
)
}

/// The validated tool call `run_internal_mcp_call` executes for `arguments`.
///
/// `arguments` is used exactly as given: it is the object the dash approval
Expand Down Expand Up @@ -1361,6 +1376,73 @@ mod tests {
assert!(text.lines().any(|line| line == version_line), "{text}");
}

/// The bridge's `engines` tool carries the same claim `rocm engines` and
/// `rocm examine` carry, so it gets the same pair of tests. These assert on
/// the serialized payload because `structuredContent.engines[].summary` is
/// the exact string an MCP client renders to the operator.
#[test]
fn engines_mcp_tool_names_vulkan_for_a_gpu_with_no_rocm_support() {
let result = internal_mcp_engines_tool_result(rocm_core::GpuSupportLevel::VulkanOnly);

let summary = engines_tool_lemonade_summary(&result);
assert!(
summary.contains("no ROCm/HIP support") && summary.contains("Vulkan llama.cpp backend"),
"the MCP client must be handed the real backend: {summary}"
);
assert!(
!summary.contains("with ROCm llama.cpp backend"),
"the stale ROCm claim must not reach an MCP client: {summary}"
);
assert!(mcp_tool_result_text(&result).contains("engine entries"));
}

#[test]
fn engines_mcp_tool_keeps_rocm_note_for_a_rocm_capable_gpu() {
let result = internal_mcp_engines_tool_result(rocm_core::GpuSupportLevel::RocmHip);

assert_eq!(
engines_tool_lemonade_summary(&result),
"default embedded Lemonade server with ROCm llama.cpp backend"
);
}

/// Pins the wiring the two tests above cannot see: that the tool reaches
/// the host-aware inventory through the live probe rather than a level
/// nailed to a constant. A hard-coded level that disagrees with this host
/// fails here.
#[test]
fn engines_mcp_tool_answers_for_the_host_it_runs_on() {
let (_root, paths) = test_paths("mcp-engines-host-probe");

let result = run_internal_mcp_call(&paths, "engines", serde_json::json!({}), false)
.expect("the engines mcp tool is read-only and should run");

assert_eq!(
result.get("isError").and_then(serde_json::Value::as_bool),
Some(false)
);
assert_eq!(
result.get("structuredContent"),
internal_mcp_engines_tool_result(host_gpu_support_level(Some(&paths)))
.get("structuredContent"),
"the `engines` tool must report this host's support level"
);
}

/// The lemonade `summary` as an MCP client reads it out of the tool result.
fn engines_tool_lemonade_summary(result: &serde_json::Value) -> String {
result
.pointer("/structuredContent/engines")
.and_then(serde_json::Value::as_array)
.expect("the engines tool must return an `engines` array")
.iter()
.find(|engine| engine.get("id").and_then(serde_json::Value::as_str) == Some("lemonade"))
.and_then(|engine| engine.get("summary"))
.and_then(serde_json::Value::as_str)
.expect("the engines tool must carry a lemonade summary")
.to_owned()
}

/// What the chat `model` tool hands back has to be what `rocm model`
/// prints, not the `--verbose` diagnostics. The verbose dump is every
/// recipe with its aliases, engine recipes, flags and policy notes; on a
Expand Down
Loading
Loading