Skip to content

test(inference): qualify llama.cpp protocol behavior #8515

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ericksoa merged 2 commits into from
Aug 7, 2026
Merged

test(inference): qualify llama.cpp protocol behavior #8515

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
9001071
test(inference): qualify llama.cpp protocol behavior
ericksoa Aug 7, 2026
f6c0039
test(inference): harden llama.cpp qualification
ericksoa Aug 7, 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
24 changes: 23 additions & 1 deletion managed-inference/images/llama-cpp/image.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -56,7 +56,29 @@ spec:
cpuFallback: reject
probes:
- health
- completion
- models
- synchronous-chat
- streaming-chat
- usage
- structured-output
- tool-call
- tool-result-continuation
- context-window
- authentication
- malformed-request
- cancellation
- client-timeout
probeBounds:
cancellationMaxTokens: 4096
clientTimeoutMilliseconds: 250
maxResponseBytes: 16777216
maxStreamEvents: 512
maxTokens:
synchronousChat: 16
streamingChat: 32
structuredOutput: 64
toolCall: 256
toolResultContinuation: 64

source:
repository: https://github.com/ggml-org/llama.cpp
Expand Down
11 changes: 10 additions & 1 deletion scripts/checks/export-llama-cpp-image-config.mts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,7 @@ import Ajv2020 from "ajv/dist/2020.js";
import YAML from "yaml";

import {
LLAMA_CPP_DGX_SPARK_PROTOCOL_PROBES,
llamaCppDgxSparkExecutionPlanSha256,
parseLlamaCppDgxSparkExecutionPlan,
} from "./llama-cpp-dgx-spark-qualification-contract.mts";
Expand Down Expand Up @@ -50,6 +51,7 @@ type ServerImageManifest = {
gpu?: unknown;
model?: unknown;
platform?: unknown;
probeBounds?: unknown;
probes?: unknown;
profile?: unknown;
recipeRef?: unknown;
Expand Down Expand Up @@ -332,6 +334,7 @@ export function loadLlamaCppImageConfig(
"gpu",
"model",
"platform",
"probeBounds",
"probes",
"profile",
"recipeRef",
Expand Down Expand Up @@ -439,7 +442,7 @@ export function loadLlamaCppImageConfig(
fullOffload: true,
vendor: "nvidia",
}) ||
JSON.stringify(qualification?.probes) !== JSON.stringify(["health", "completion"])
JSON.stringify(qualification?.probes) !== JSON.stringify(LLAMA_CPP_DGX_SPARK_PROTOCOL_PROBES)
) {
throw new Error("invalid llama.cpp image publication contract");
}
Expand Down Expand Up @@ -536,6 +539,7 @@ export function loadLlamaCppImageConfig(
id: qualificationModel?.id,
},
platform: qualification?.platform,
probeBounds: qualification?.probeBounds,
probes: qualification?.probes,
profile: qualification?.profile,
recipeRef: qualificationRecipeRef,
Expand Down Expand Up @@ -669,7 +673,12 @@ export function loadLlamaCppImageConfig(
revision: spec?.source?.revision,
},
},
qualification: {
probeBounds: qualification?.probeBounds,
probes: qualification?.probes,
},
recipe: {
capabilities: recipe.spec.capabilities,
id: recipe.metadata.id,
model: {
acquisition: {
Expand Down
Loading
Loading