Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-20 01:31:42 +02:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'upstream' into concedo_experimental

Browse Source 
# Conflicts:
#	.github/workflows/release.yml
#	.github/workflows/server.yml
#	examples/model-conversion/requirements.txt
#	examples/model-conversion/scripts/causal/run-casual-gen-embeddings-org.py
#	examples/speculative-simple/README.md
#	examples/speculative-simple/speculative-simple.cpp
#	ggml/src/ggml-opencl/ggml-opencl.cpp
#	requirements/requirements-convert_hf_to_gguf.txt
#	requirements/requirements-convert_lora_to_gguf.txt
#	scripts/hip/gcn-cdna-vgpr-check.py
#	tests/test-backend-ops.cpp
#	tests/test-chat.cpp
#	tools/imatrix/imatrix.cpp
#	tools/mtmd/CMakeLists.txt
This commit is contained in:
Concedo
2026-08-12 21:34:23 +08:00
parent ddd2691551 4dd127584b
commit f556b13a4e
79 changed files with 3714 additions and 282 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/ui/tests/client/README-perf.md
+6 -6
View File
@@ -33,12 +33,12 @@ npx vitest --project=client --run tests/client/agentic-stream.perf.svelte.test.t
The point of the harness is the _scaling curve_, not any single number.
| Knob | Reads on |
| --------------------------- | ---------------------------------------------------------------------------------------------------- |
| `priorToolCalls` (0/1/5/20) | the reactive fan-out. Flat => no fan-out. Linear => confirmed. |
| `toolResultBytes` | whole-blob string scans (`extractSearchResults`, `parseToolResultWithImages`, `classifyToolResult`). |
| `editFileEdits` | `computeLineDiff`, the O(m\*n) LCS. |
| `openCodeFence` | `hljs.highlightAuto` on partial code. |
| Knob | Reads on |
| --------------------------- | --------------------------------------------------------------------------------------------------- |
| `priorToolCalls` (0/1/5/20) | the reactive fan-out. Flat => no fan-out. Linear => confirmed. |
| `toolResultBytes` | whole-blob string scans (`extractSearchResults`, `parseToolResultWithMedia`, `classifyToolResult`). |
| `editFileEdits` | `computeLineDiff`, the O(m\*n) LCS. |
| `openCodeFence` | `hljs.highlightAuto` on partial code. |
Deliberately no hard assertions: CI timing is noisy and the value here is the
before/after delta, not a gate.
tools/ui/tests/unit/agentic-hotpath.bench.ts
+5 -5
View File
@@ -6,7 +6,7 @@
//
// Run: npx vitest bench --project=unit tests/unit/agentic-hotpath.bench.ts
import { classifyToolResult, parseToolResultWithImages } from '$lib/utils/agentic';
import { classifyToolResult, parseToolResultWithMedia } from '$lib/utils/agentic';
import { detectIncompleteCodeBlock, highlightCode } from '$lib/utils/code';
import { computeLineDiff } from '$lib/utils/compute-line-diff';
import { preprocessLaTeX } from '$lib/utils/latex-protection';
@@ -200,17 +200,17 @@ describe('exit-code regex', () => {
// --- per-line result parsers ----------------------------------------------
describe('parseToolResultWithImages', () => {
describe('parseToolResultWithMedia', () => {
bench('1KB', () => {
parseToolResultWithImages(SHELL_OUTPUT_1KB, []);
parseToolResultWithMedia(SHELL_OUTPUT_1KB, []);
});
bench('200KB', () => {
parseToolResultWithImages(SHELL_OUTPUT_200KB, []);
parseToolResultWithMedia(SHELL_OUTPUT_200KB, []);
});
bench('2MB', () => {
parseToolResultWithImages(SHELL_OUTPUT_2MB, []);
parseToolResultWithMedia(SHELL_OUTPUT_2MB, []);
});
});