Merge pull request 'udpate chromadb import to be lazy' (#27) from fix-random-illegal-instruction into master

Reviewed-on: #27

src/ai_generators/blog_flow.py

@@ -17,14 +17,19 @@ journalist draft is chunked, embedded, and stored in a collection; the editor
 receives the top-N most relevant chunks as context.
 """
 
+from __future__ import annotations
+
 import json
 import os
 import random
 import re
 import string
 from datetime import datetime
+from typing import TYPE_CHECKING
+
+if TYPE_CHECKING:
+    import chromadb  # noqa: F811
 
-import chromadb
 from crewai.flow.flow import Flow, listen, start
 from ollama import Client
 from pydantic import BaseModel, ConfigDict
@@ -85,7 +90,21 @@ class BlogFlow(Flow[BlogFlowState]):
         )
 
     @staticmethod
-    def _get_chroma_client() -> chromadb.HttpClient:
+    def _get_chroma_client() -> "chromadb.HttpClient":
+        # Lazily import chromadb here rather than at module level.
+        # chromadb unconditionally loads hnswlib (a native C++ library
+        # compiled with AVX instructions) even when using HttpClient
+        # to talk to an external ChromaDB server.  Deferring the import
+        # avoids "Illegal instruction" (SIGILL) crashes in environments
+        # that lack AVX support (e.g. act, older CI runners).
+        try:
+            import chromadb
+        except ImportError as exc:
+            raise RuntimeError(
+                "chromadb is required for the editor phase but could not be "
+                f"imported: {exc}"
+            ) from exc
+
         chroma_port = int(os.environ["CHROMA_PORT"])
         return chromadb.HttpClient(host=os.environ["CHROMA_HOST"], port=chroma_port)
 
@@ -127,7 +146,7 @@ class BlogFlow(Flow[BlogFlowState]):
             print(f"Error generating embeddings: {exc}")
             return []
 
-    def _load_drafts_to_vector_db(self, drafts: list[str]) -> chromadb.Collection:
+    def _load_drafts_to_vector_db(self, drafts: list[str]) -> "chromadb.Collection":
         """Load journalist drafts into a new ChromaDB collection and return it."""
         chroma = self._get_chroma_client()
         collection_name = (
@@ -165,7 +184,7 @@ class BlogFlow(Flow[BlogFlowState]):
         return collection
 
     @staticmethod
-    def _query_vector_db(collection: chromadb.Collection, query_text: str) -> str:
+    def _query_vector_db(collection: "chromadb.Collection", query_text: str) -> str:
         """Query the ChromaDB collection and return the most relevant
         document chunks joined as a single string."""
         ollama_client = BlogFlow._get_ollama_client()