From 36aa6aef1005ae95158f9103e02f76ec0f919512 Mon Sep 17 00:00:00 2001 From: JerrettDavis Date: Sun, 26 Apr 2026 20:37:14 -0500 Subject: [PATCH 1/2] fix(ci): use llama3.2:1b model in integration workflow for reliability Replace qwen2.5:0.5b with llama3.2:1b as the Ollama model in the integration workflow. The smaller model is too weak/flaky for ProviderFeatureIntegrationTests.Ollama_WhenAvailable_BuildsKernel_AndCompletesToolTurn. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> --- .github/workflows/integration.yml | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/.github/workflows/integration.yml b/.github/workflows/integration.yml index 6c4a75ad..60004c1c 100644 --- a/.github/workflows/integration.yml +++ b/.github/workflows/integration.yml @@ -10,7 +10,7 @@ on: chat_model: description: 'Ollama chat model to use' required: false - default: 'qwen2.5:0.5b' + default: 'llama3.2:1b' embedding_model: description: 'Ollama embedding model to use' required: false @@ -18,7 +18,7 @@ on: env: DOTNET_NOLOGO: true - OLLAMA_CHAT_MODEL: ${{ inputs.chat_model || 'qwen2.5:0.5b' }} + OLLAMA_CHAT_MODEL: ${{ inputs.chat_model || 'llama3.2:1b' }} OLLAMA_EMBEDDING_MODEL: ${{ inputs.embedding_model || 'all-minilm:22m' }} EXTENSIONS_INTEGRATION_TESTS: 'true' TUI_INTEGRATION_TESTS: 'true' From e356a0625da3812d00b2cbf256295746c2df8f1c Mon Sep 17 00:00:00 2001 From: "github-actions[bot]" Date: Mon, 27 Apr 2026 01:49:38 +0000 Subject: [PATCH 2/2] ci: retrain intent classifier model Generated 50 examples via llama3.2:1b Triggered by: push Workflow run: 24972732230 --- .../Models/intent_classifier.zip | Bin 22534 -> 22485 bytes .../Models/ollama_training_data.jsonl | 22 +++++++++--------- 2 files changed, 11 insertions(+), 11 deletions(-) diff --git a/src/JD.AI.Workflows/Models/intent_classifier.zip b/src/JD.AI.Workflows/Models/intent_classifier.zip index 187ed4de3b0381bed55d3f51b0b8b1564b560dba..2a19992c16fb7a0de00daedbfcd930d508988592 100644 GIT binary patch delta 834 zcmZqMz<6~%BTs-gGm8iV0|N(xD&It&WF}R~TBjRT&{)5fxm^TpZ4{tp6een?HTo2cGzdm?Y~^Lc;L>(V6t)u%0S=yTur z;YC8`+dkHB|9)+M`R!eg>Mtd3}fuj>&t&MCC!IM_;M+ z*eAunuuqzSK?`OykY3VwV)Dl@OOQsLa2cj^-VlxKlMBKHz|MN&s_48MsBjN3odNa3 x!IH-3J`)>-CvOfH1!+DNE~9YEmw_QTIU_YUF~FOZ4Wv>62unE`7`*&IJOF+h4AlSt delta 860 zcmcb*p0RBMBTs-gGm8iV0|N&`2IoYcWTp(xjfHlMKx*?wMqd!UiTMvRi2ILCmkmVg z@zhpb9eq*nQKzBin31_e44xD+6KSca;f@Y?V>JAjd%S& z<)<|M?0xS>&a3I3h#>1Or4Da7PFg=sXBI(AyBrSpMkt=?|X^Yrkai zrLX$>@$lbk@ijYlR3#bRtdXC4y*@+u(IkP(>S_wE$K4iPRy(SA=&<6!#m-I69xeV( z^SO#;M5{e~S~?rgebO&>ZSIlNl;pHqR+PR(YxQyw_LP#BEA791(6U_Ye1>zMS4m;y z;_ic7jEs76rZ%CgB|KK&+%)lFSe=t?#-w5J;g-{g%Wgy-M2$c~CR&ohK*#v>*wgiC% z<$}Rf2ZZ_rp<+U!K-zbNgz%w8^<=5g2ziju&-x$67Qj?!2E=wi3C|qX*Lgq?a@@Og<2%1WrFaP_@xlYCZNzF)-|t zW?;}lQTr39))HjF?=Tsr|K8A)$2vK|UjQT;7%syk;XAQWcyd#?C`e#txQqg~9|J>h baz<)yVt_X*8^|~bAS~r%V5kPlFfafBLVhFd diff --git a/src/JD.AI.Workflows/Models/ollama_training_data.jsonl b/src/JD.AI.Workflows/Models/ollama_training_data.jsonl index dddcf6d0..1dad6dd1 100644 --- a/src/JD.AI.Workflows/Models/ollama_training_data.jsonl +++ b/src/JD.AI.Workflows/Models/ollama_training_data.jsonl @@ -3,16 +3,16 @@ {"Prompt":"Archive old npm logs to S3","IsWorkflow":false} {"Prompt":"What would make python more maintainable?","IsWorkflow":false} {"Prompt":"Create a worktree for the python feature branch","IsWorkflow":false} -{"Prompt":"Deploy database to the staging environment","IsWorkflow":false} -{"Prompt":"Scale the git service to 3 replicas","IsWorkflow":true} +{"Prompt":"Deploy database to the staging environment","IsWorkflow":true} +{"Prompt":"Scale the git service to 3 replicas","IsWorkflow":false} {"Prompt":"Background: cloud has been causing issues since last week","IsWorkflow":false} -{"Prompt":"Does cloud support WebSocket connections?","IsWorkflow":false} +{"Prompt":"Does cloud support WebSocket connections?","IsWorkflow":true} {"Prompt":"Add fuzzing tests for the database input handler","IsWorkflow":false} {"Prompt":"Seed the cloud database with test fixtures","IsWorkflow":false} {"Prompt":"How do I configure ci-cd authentication?","IsWorkflow":false} {"Prompt":"Replace all TODO comments in the file-ops codebase with proper issues","IsWorkflow":false} {"Prompt":"What do you think about networking for this use case?","IsWorkflow":false} -{"Prompt":"Write a Dockerfile for the monitoring service","IsWorkflow":true} +{"Prompt":"Write a Dockerfile for the monitoring service","IsWorkflow":false} {"Prompt":"How can we improve the api developer experience?","IsWorkflow":false} {"Prompt":"What\u0027s new in the devops release notes?","IsWorkflow":false} {"Prompt":"Add a required reviewer to the shell PR pipeline","IsWorkflow":false} @@ -28,23 +28,23 @@ {"Prompt":"Update the CHANGELOG for git","IsWorkflow":false} {"Prompt":"Background: database has been causing issues since last week","IsWorkflow":false} {"Prompt":"Create a PR that adds telemetry and add reviewers","IsWorkflow":false} -{"Prompt":"Remove the dead code from the devops module","IsWorkflow":true} +{"Prompt":"Remove the dead code from the devops module","IsWorkflow":false} {"Prompt":"Should we use git or stick with the current approach?","IsWorkflow":false} {"Prompt":"Squash the last 5 commits and force push","IsWorkflow":false} -{"Prompt":"Update the CHANGELOG for git","IsWorkflow":true} +{"Prompt":"Update the CHANGELOG for git","IsWorkflow":false} {"Prompt":"Replace all TODO comments in the networking codebase with proper issues","IsWorkflow":false} {"Prompt":"Write integration tests for the shell API","IsWorkflow":false} -{"Prompt":"Does debugging support WebSocket connections?","IsWorkflow":true} +{"Prompt":"Does debugging support WebSocket connections?","IsWorkflow":false} {"Prompt":"What patterns work best for docker-compose at scale?","IsWorkflow":false} {"Prompt":"Fix the formatting in all database source files","IsWorkflow":false} -{"Prompt":"Tag the release v1.1.0 for file-ops","IsWorkflow":true} +{"Prompt":"Tag the release v1.1.0 for file-ops","IsWorkflow":false} {"Prompt":"Run a migration to add index on docs.events","IsWorkflow":true} -{"Prompt":"Build and push the testing image to the registry","IsWorkflow":true} +{"Prompt":"Build and push the testing image to the registry","IsWorkflow":false} {"Prompt":"Write integration tests for the testing API","IsWorkflow":false} {"Prompt":"Audit the npm packages in cloud for vulnerabilities","IsWorkflow":false} {"Prompt":"Background: networking has been causing issues since last week","IsWorkflow":false} {"Prompt":"Prune all stopped containers and dangling images","IsWorkflow":false} -{"Prompt":"Does code-review support WebSocket connections?","IsWorkflow":true} +{"Prompt":"Does code-review support WebSocket connections?","IsWorkflow":false} {"Prompt":"Check the health of devops in production","IsWorkflow":false} {"Prompt":"Dump the database production DB to a file","IsWorkflow":true} -{"Prompt":"Should we use git or stick with the current approach?","IsWorkflow":true} +{"Prompt":"Should we use git or stick with the current approach?","IsWorkflow":false}