Issue fixes

MemMachine · Tianyang-Zhang · Mar 3, 2026 · Mar 6, 2026 · Mar 8, 2026 · Mar 8, 2026
commit bf4ecc661761aee9943412c34128c33b463673e0
diff --git a/.github/workflows/installation-test.yml b/.github/workflows/installation-test.yml
@@ -73,8 +73,9 @@ jobs:
         shell: bash
         run: |
           set -eo pipefail
-          whl_name=$(ls *server*.whl)
-          python -m pip install --find-links . "$whl_name"
+          common_whl=$(ls memmachine_common-*.whl | head -1)
+          server_whl=$(ls memmachine_server-*.whl | head -1)
+          python -m pip install --find-links . "$common_whl" "$server_whl"
 
       #
       # ─────────────────────────────────────
@@ -105,8 +106,9 @@ jobs:
         run: |
           set -eo pipefail
           export PYTHONUTF8=1
-          whl_name=$(ls *server*.whl)
-          python -m pip install --find-links . "$whl_name"
+          common_whl=$(ls memmachine_common-*.whl | head -1)
+          server_whl=$(ls memmachine_server-*.whl | head -1)
+          python -m pip install --find-links . "$common_whl" "$server_whl"
 
       #
       # ─────────────────────────────────────
@@ -124,8 +126,9 @@ jobs:
         shell: bash
         run: |
           set -eo pipefail
-          whl_name=$(ls *server*.whl)
-          python -m pip install --find-links . "$whl_name"
+          common_whl=$(ls memmachine_common-*.whl | head -1)
+          server_whl=$(ls memmachine_server-*.whl | head -1)
+          python -m pip install --find-links . "$common_whl" "$server_whl"
 
       #
       # ─────────────────────────────────────

diff --git a/.github/workflows/test-server-package.yml b/.github/workflows/test-server-package.yml
@@ -52,8 +52,9 @@ jobs:
         shell: bash
         run: |
           set -eo pipefail
-          whl_name=$(ls dist/memmachine_server-*.whl | head -1)
-          pip install --find-links dist/ "$whl_name"
+          common_whl=$(ls dist/memmachine_common-*.whl | head -1)
+          server_whl=$(ls dist/memmachine_server-*.whl | head -1)
+          pip install --find-links dist/ "$common_whl" "$server_whl"
 
       - name: Test server package imports
         shell: bash

diff --git a/evaluation/retrieval_skill/benchmark_config.py b/evaluation/retrieval_skill/benchmark_config.py
@@ -16,14 +16,12 @@
 
 from __future__ import annotations
 
-import json
 import os
 import re
-from dataclasses import dataclass, field
+from dataclasses import dataclass
 from pathlib import Path
 from typing import Any
 
-
 # ---------------------------------------------------------------------------
 # Environment-variable substitution
 # ---------------------------------------------------------------------------

diff --git a/evaluation/retrieval_skill/llm_judge.py b/evaluation/retrieval_skill/llm_judge.py
@@ -44,7 +44,12 @@
 def _get_openai_client() -> openai.OpenAI:
     global _openai_client
     if _openai_client is None:
-        _openai_client = openai.OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
+        api_key = os.getenv("OPENAI_API_KEY")
+        if not api_key:
+            raise RuntimeError(
+                "OPENAI_API_KEY must be set when no eval_llm client is provided."
+            )
+        _openai_client = openai.OpenAI(api_key=api_key)
     return _openai_client
 
 
@@ -147,7 +152,10 @@ def main():
                 continue
 
             label = evaluate_llm_judge(
-                question, gold_answer, generated_answer, args.model,
+                question,
+                gold_answer,
+                generated_answer,
+                args.model,
                 eval_llm=eval_llm,
             )
             LLM_JUDGE[category].append(label)

diff --git a/evaluation/retrieval_skill/locomo_search.py b/evaluation/retrieval_skill/locomo_search.py
@@ -7,12 +7,13 @@
 from typing import Any
 
 from dotenv import load_dotenv
-from memmachine_server.common.utils import async_with
 
 REPO_ROOT = Path(__file__).resolve().parents[2]
 if str(REPO_ROOT) not in sys.path:
     sys.path.append(str(REPO_ROOT))
 
+from memmachine_server.common.utils import async_with  # noqa: E402
+
 from evaluation.utils import skill_utils  # noqa: E402
 
 ANSWER_PROMPT = """
@@ -157,7 +158,9 @@ async def run_locomo(  # noqa: C901
                 speaker = message["speaker"]
                 full_content.append(f"[{session_datetime}] {speaker}: {text}")
 
-        answer_model_name = answer_llm.model_name if answer_llm is not None else "gpt-5-mini"
+        answer_model_name = (
+            answer_llm.model_name if answer_llm is not None else "gpt-5-mini"
+        )
         _, model, query_skill = await skill_utils.init_memmachine_params(
             vector_graph_store=vector_graph_store,
             session_id=group_id,