updating episodic memory prompts (#11)

* updating episodic memory prompts

* code improvements

* code improvements

---------

Co-authored-by: Aayush Kataria <aayushkataria@Aayushs-MacBook-Pro-2.local>

Author: aayush3011

Docs/azure_testing.md

@@ -314,9 +314,9 @@ for i in range(10):
 
 # Wait for the change-feed processor to catch up, then read derived memories.
 import time; time.sleep(15)
-print(memory.get_memories(user_id="user-1", thread_id="thread-1", memory_type="summary"))
-print(memory.get_memories(user_id="user-1", memory_type="fact"))
-print(memory.get_memories(user_id="user-1", memory_type="user_summary"))
+print(memory.get_memories(user_id="user-1", thread_id="thread-1", memory_types=["summary"]))
+print(memory.get_memories(user_id="user-1", memory_types=["fact"]))
+print(memory.get_memories(user_id="user-1", memory_types=["user_summary"]))
 ```
 
 ### Change feed auto-processing
@@ -339,7 +339,7 @@ for i in range(3):
 # Wait a few seconds for the change feed to trigger, then check:
 import time
 time.sleep(10)
-results = memory.get_memories(user_id="user-1", thread_id=thread_id, memory_type="summary")
+results = memory.get_memories(user_id="user-1", thread_id=thread_id, memory_types=["summary"])
 print(results)  # Should contain an auto-generated summary
 ```
 
@@ -348,9 +348,9 @@ Check the Function App logs to confirm the `on_memory_change` trigger fired and
 ### Verify stored results
 
 ```python
-print(memory.get_memories(user_id="user-1", memory_type="summary"))
-print(memory.get_memories(user_id="user-1", memory_type="fact"))
-print(memory.get_memories(user_id="user-1", memory_type="user_summary"))
+print(memory.get_memories(user_id="user-1", memory_types=["summary"]))
+print(memory.get_memories(user_id="user-1", memory_types=["fact"]))
+print(memory.get_memories(user_id="user-1", memory_types=["user_summary"]))
 ```
 
 ---

Docs/design_patterns.md

@@ -162,7 +162,7 @@ Search across all memories (or filter by type) to find the most relevant context
 facts = await mem.search_cosmos(
     search_terms="database migration requirements",
     user_id="user-1",
-    memory_type="fact",
+    memory_types=["fact"],
     top_k=10,
 )
 
@@ -185,10 +185,10 @@ profile = await mem.get_user_summary(user_id="user-1")
 
 ```python
 # All summaries for a user
-summaries = await mem.get_memories(user_id="user-1", memory_type="summary")
+summaries = await mem.get_memories(user_id="user-1", memory_types=["summary"])
 
 # All facts
-facts = await mem.get_memories(user_id="user-1", memory_type="fact")
+facts = await mem.get_memories(user_id="user-1", memory_types=["fact"])
 
 # Filter by thread_id
 thread_turns = await mem.get_memories(user_id="user-1", thread_id=THREAD_ID)
@@ -204,7 +204,7 @@ A typical chat application lifecycle looks like this:
 New session starts
   │
   ├─ Retrieve user summary          (get_user_summary)
-  ├─ Semantic search for prior facts (search_cosmos, type="fact")
+  ├─ Semantic search for prior facts (search_cosmos, memory_types=["fact"])
   │
   │  ┌── Conversation loop ──┐
   │  │ Store each turn        │  (add_cosmos)
@@ -221,7 +221,7 @@ New session starts
 ```python
 # --- Session start ---
 profile = await mem.get_user_summary(user_id="user-1")
-relevant = await mem.search_cosmos("topic of interest", user_id="user-1", memory_type="fact", top_k=5)
+relevant = await mem.search_cosmos("topic of interest", user_id="user-1", memory_types=["fact"], top_k=5)
 
 # Build system prompt with profile and relevant facts
 system_prompt = build_prompt(profile, relevant)
@@ -280,7 +280,7 @@ await mem.extract_facts(user_id="user-1", thread_id="research-thread")
 facts = await mem.search_cosmos(
     search_terms="source database schema foreign keys",
     user_id="user-1",
-    memory_type="fact",
+    memory_types=["fact"],
     top_k=10,
 )
 

Docs/local_testing.md

@@ -377,7 +377,7 @@ for i in range(3):
 5. Verify the summary was created:
 
 ```python
-result = memory.get_memories(user_id="user-001", thread_id=thread_id, memory_type="summary")
+result = memory.get_memories(user_id="user-001", thread_id=thread_id, memory_types=["summary"])
 print(result)
 ```
 

README.md

@@ -202,7 +202,7 @@ Filter at retrieval time:
 
 ```python
 results = memory.search_cosmos("user preferences", user_id="u1", min_confidence=0.7)
-high_conf_facts = memory.get_memories(user_id="u1", memory_type="fact", min_confidence=0.7)
+high_conf_facts = memory.get_memories(user_id="u1", memory_types=["fact"], min_confidence=0.7)
 ```
 
 ### Memory Reconciliation

Samples/Demo.ipynb

@@ -510,7 +510,7 @@
     "    print(f\"  [{m['id'][:8]}...] {m['content'][:60]}\")\n",
     "\n",
     "# Filter by type\n",
-    "facts = memory.get_local(memory_type=\"fact\")\n",
+    "facts = memory.get_local(memory_types=[\"fact\"])\n",
     "print(f\"\\nFact memories: {len(facts)}\")\n",
     "for m in facts:\n",
     "    print(f\"  [{m['id'][:8]}...] {m['content']}\")\n",
@@ -2739,7 +2739,7 @@
    "source": [
     "# Inspect the persisted derived memories across BOTH threads for this user.\n",
     "for mt in (\"fact\", \"episodic\", \"procedural\"):\n",
-    "    docs = memory.get_memories(user_id=USER_ID, memory_type=mt)\n",
+    "    docs = memory.get_memories(user_id=USER_ID, memory_types=[mt])\n",
     "    print(f\"\\n{mt.upper()}S ({len(docs)}):\")\n",
     "    for d in docs:\n",
     "        print(f\"  • {d['content'][:100]}  [salience={d.get('salience')}]\")\n"
@@ -2879,7 +2879,7 @@
     "stored = memory.get_user_summary(user_id=user_id)\n",
     "if stored:\n",
     "    print(\"User Summary for\", user_id)\n",
-    "    print(stored[0][\"content\"])\n",
+    "    print(stored[\"content\"])\n",
     "else:\n",
     "    print(\"No user summary found — run the generate_user_summary cell first.\")\n"
    ]

Samples/Demo_async.ipynb

@@ -499,7 +499,7 @@
     "    print(f\"  [{m['id'][:8]}...] {m['content'][:60]}\")\n",
     "\n",
     "# Filter by type\n",
-    "facts = memory.get_local(memory_type=\"fact\")\n",
+    "facts = memory.get_local(memory_types=[\"fact\"])\n",
     "print(f\"\\nFact memories: {len(facts)}\")\n",
     "for m in facts:\n",
     "    print(f\"  [{m['id'][:8]}...] {m['content']}\")\n",
@@ -2631,7 +2631,7 @@
    "source": [
     "# Inspect the persisted derived memories across BOTH threads for this user.\n",
     "for mt in (\"fact\", \"episodic\", \"procedural\"):\n",
-    "    docs = await memory.get_memories(user_id=USER_ID, memory_type=mt)\n",
+    "    docs = await memory.get_memories(user_id=USER_ID, memory_types=[mt])\n",
     "    print(f\"\\n{mt.upper()}S ({len(docs)}):\")\n",
     "    for d in docs:\n",
     "        print(f\"  • {d['content'][:100]}  [salience={d.get('salience')}]\")\n"
@@ -2765,7 +2765,7 @@
     "stored = await memory.get_user_summary(user_id=user_id)\n",
     "if stored:\n",
     "    print(\"User Summary for\", user_id)\n",
-    "    print(stored[0][\"content\"])\n",
+    "    print(stored[\"content\"])\n",
     "else:\n",
     "    print(\"No user summary found — run the generate_user_summary cell first.\")\n"
    ]

Samples/Demo_function_app.ipynb

@@ -254,7 +254,7 @@
     "## 5. Wait for the Function App to produce a summary\n",
     "\n",
     "`process_now_and_wait` polls Cosmos for the thread's summary doc until `timeout` seconds. This is **RU-costly**\n",
-    "(repeated `get_memories(memory_type='summary')` queries) — only use it for demos and tests.\n",
+    "(repeated `get_memories(memory_types=['summary'])` queries) — only use it for demos and tests.\n",
     "\n",
     "> If this returns `False`, check that:\n",
     "> * The Function App is deployed and running.\n",
@@ -594,7 +594,7 @@
     "def _show():\n",
     "    counts = {}\n",
     "    for mt in (\"summary\", \"fact\", \"episodic\", \"procedural\", \"user_summary\"):\n",
-    "        docs = memory.get_memories(user_id=USER_ID, memory_type=mt)\n",
+    "        docs = memory.get_memories(user_id=USER_ID, memory_types=[mt])\n",
     "        counts[mt] = len(docs)\n",
     "        print(f\"\\n{mt.upper()}S ({len(docs)}):\")\n",
     "        for d in docs:\n",
@@ -610,7 +610,7 @@
     "        break\n",
     "    print(f\"\\n... fact / procedural extraction still in flight (attempt {attempt+1}/6, sleeping 15s)\")\n",
     "    time.sleep(15)\n",
-    "    counts = {mt: len(memory.get_memories(user_id=USER_ID, memory_type=mt))\n",
+    "    counts = {mt: len(memory.get_memories(user_id=USER_ID, memory_types=[mt]))\n",
     "              for mt in (\"summary\", \"fact\", \"episodic\", \"procedural\", \"user_summary\")}\n",
     "\n",
     "print(\"\\n=== Final state ===\")\n",
@@ -627,7 +627,7 @@
     "* **Per-turn embedding**: the SDK auto-embeds non-`turn` documents you `add_cosmos` directly. Raw `turn`\n",
     "  records are intentionally not embedded — the function app does that during summary/extraction.\n",
     "* **Search across function-app-produced memories**: works exactly as in the in-process demo:\n",
-    "  `memory.search_cosmos(search_terms=\"…\", memory_type=\"fact\", user_id=USER_ID)`.\n",
+    "  `memory.search_cosmos(search_terms=\"…\", memory_types=[\"fact\"], user_id=USER_ID)`.\n",
     "* **Switch back to in-process** for ad-hoc work: instantiate the client without the `processor=` kwarg\n",
     "  (or pass `InProcessProcessor()` explicitly) and use `generate_thread_summary` / `extract_memories`\n",
     "  directly."

Samples/Demo_function_app_async.ipynb

@@ -537,7 +537,7 @@
     "async def _show():\n",
     "    counts = {}\n",
     "    for mt in (\"summary\", \"fact\", \"episodic\", \"procedural\", \"user_summary\"):\n",
-    "        docs = await memory.get_memories(user_id=USER_ID, memory_type=mt)\n",
+    "        docs = await memory.get_memories(user_id=USER_ID, memory_types=[mt])\n",
     "        counts[mt] = len(docs)\n",
     "        print(f\"\\n{mt.upper()}S ({len(docs)}):\")\n",
     "        for d in docs:\n",
@@ -554,7 +554,7 @@
     "    await asyncio.sleep(15)\n",
     "    counts = {}\n",
     "    for mt in (\"summary\", \"fact\", \"episodic\", \"procedural\", \"user_summary\"):\n",
-    "        docs = await memory.get_memories(user_id=USER_ID, memory_type=mt)\n",
+    "        docs = await memory.get_memories(user_id=USER_ID, memory_types=[mt])\n",
     "        counts[mt] = len(docs)\n",
     "\n",
     "print(\"\\n=== Final state ===\")\n",

Samples/advanced_memory_lifecycle.py

@@ -92,7 +92,7 @@ def main() -> None:
 
     _header(5, "Archive: delete raw turns, keep derived memories")
     deleted = 0
-    for m in mem.get_memories(user_id=user_id, thread_id=thread_id, memory_type="turn"):
+    for m in mem.get_memories(user_id=user_id, thread_id=thread_id, memory_types=["turn"]):
         mem.delete_cosmos(memory_id=m["id"], thread_id=thread_id, user_id=user_id)
         deleted += 1
     print(f"  deleted {deleted} raw turn(s)")

Samples/processing_fact_extraction.py

@@ -75,7 +75,7 @@ def main() -> None:
     print(f"Extraction stats: {json.dumps(stats, indent=2)}\n")
 
     for kind in ("fact", "procedural", "episodic"):
-        items = mem.get_memories(user_id=user_id, memory_type=kind)
+        items = mem.get_memories(user_id=user_id, memory_types=[kind])
         print(f"{kind.upper()}S ({len(items)}):")
         for it in items:
             tags = ", ".join(it.get("tags") or [])