add searching for different stuff

database.py

···
                 score=1.0,
             )
 
+    def get_avatar_by_did(self, did: str) -> Optional[ResultWithVector]:
+        result = self.client.scroll(
+            collection_name=self.avatar_collection_name,
+            scroll_filter=Filter(
+                must=[FieldCondition(key="did", match=MatchValue(value=did))]
+            ),
+            with_vectors=True,
+            with_payload=True,
+        )
+
+        if result and result[0] and len(result[0]) > 0:
+            point = result[0][0]
+            return ResultWithVector(
+                did=point.payload["did"],
+                payload=point.payload,
+                vector=point.vector,
+                score=1.0,
+            )
+
 
 QDRANT_SERVICE = QdrantService()

search.py

···
 import logging
+import math
 import sys
 from typing import List
 import click
···
 from config import CONFIG
 from database import QDRANT_SERVICE, Result
 from embedder import EMBEDDING_SERVICE
+from retina import RETINA_CLIENT, binary_to_float_vector, hex_to_binary
 
 
 logging.basicConfig(
···
             more = None
             if type == "profile":
                 more = result.payload.get("description")
+            elif type == "avatar":
+                cid = result.payload.get("cid")
+                more = f"https://cdn.bsky.app/img/feed_thumbnail/plain/{result.did}/{cid}@jpeg"
+            elif type == "post":
+                more = result.payload.get("text")
 
             if more is not None:
                 row.append(more)
···
     show_more: bool,
 ):
     # TODO: would be nice if these were flags instead
-    if type != "profile":
+    if type not in ["profile", "avatar", "post"]:
         raise Exception("invalid type")
 
     QDRANT_SERVICE.initialize()
···
 
             results = QDRANT_SERVICE.search_similar(
                 collection_name=CONFIG.qdrant_profile_collection_name,
+                query_vector=query_vector,
+                limit=limit,
+                score_threshold=math.sqrt(threshold),
+            )
+
+            display_results(type, description, results, show_more)
+        elif type == "avatar":
+            if not query:
+                console.print("[cyan]Looking up avatar...[/cyan]")
+                avatar = QDRANT_SERVICE.get_avatar_by_did(did)
+
+                if not avatar:
+                    console.print(f"[red]Avatar not found: {did}[/red]")
+                    sys.exit(1)
+
+                cid = avatar.payload.get("cid")
+                query_vector = avatar.vector
+            else:
+                pts = query.split("/")
+
+                if len(pts) != 8:
+                    console.print("[red]Invalid avatar URL provided[/red]")
+                    sys.exit(1)
+
+                did = pts[6]
+                cid = pts[7].split("@")[0]
+
+                resp = RETINA_CLIENT.get_image_hash(did, cid)
+
+                if resp.quality_too_low or resp.hash is None:
+                    console.print("[red]Hash quality too low[/red]")
+                    sys.exit(1)
+
+                query_vector = binary_to_float_vector(hex_to_binary(resp.hash))
+
+            console.print("[cyan]Looking up similar avatars...[/cyan]")
+
+            results = QDRANT_SERVICE.search_similar(
+                collection_name=CONFIG.qdrant_avatar_collection_name,
+                query_vector=query_vector,
+                limit=limit,
+                score_threshold=threshold,
+            )
+
+            display_results(type, cid, results, show_more)
+        elif type == "post":
+            if not query:
+                console.print("[red]Must supply input for post search[/red]")
+                sys.exit(1)
+            else:
+                EMBEDDING_SERVICE.initialize()
+
+                description = query
+                query_vector = EMBEDDING_SERVICE.encode(query)
+
+            console.print("[cyan]Looking up similar posts...[/cyan]")
+
+            results = QDRANT_SERVICE.search_similar(
+                collection_name=CONFIG.qdrant_post_collection_name,
                 query_vector=query_vector,
                 limit=limit,
                 score_threshold=threshold,