examples and applications

examples_and_applications/healthcare_conversation_rag/create_dataset.py

@@ -0,0 +1,100 @@
+from datasets import load_dataset
+from datetime import date
+
+# ── config ────────────────────────────────────────────────────────────────────
+REPO_ID     = "pavanmantha/doctor_patient_conversation"
+COLLECTION  = "doctor_patient_conversation"
+OUTPUT_FILE = "data_sets/source_data.qql"
+TODAY       = date.today().isoformat()   # e.g. 2026-05-16
+BATCH_SIZE  = 200                        # rows per INSERT block (stays under 33MB)
+# ─────────────────────────────────────────────────────────────────────────────
+
+
+def escape(text: str) -> str:
+    """Escape single-quotes inside field values."""
+    return text.replace("\\", "\\\\").replace("'", "\\'")
+
+
+def build_record(row: dict) -> str:
+    description = escape((row.get("description") or "").strip())
+    text        = escape((row.get("conversation") or "").strip())
+    status      = escape((row.get("status")       or "").strip())
+
+    return (
+        "  {\n"
+        f"    'description': '{description}',\n"
+        f"    'text': '{text}',\n"
+        f"    'status': '{status}'\n"
+        "  }"
+    )
+
+
+def write_batch(f, batch: list, batch_num: int, collection: str):
+    f.write(f"\n-- Batch {batch_num} ({len(batch)} records)\n")
+    f.write(f"INSERT BULK INTO COLLECTION {collection} VALUES [\n")
+    for i, record in enumerate(batch):
+        is_last = (i == len(batch) - 1)
+        f.write(record)
+        f.write("\n" if is_last else ",\n")
+    f.write("]\n")
+
+
+def main():
+    print(f"Loading dataset from '{REPO_ID}' ...")
+    ds = load_dataset(REPO_ID, split="train")
+    total = len(ds)
+    num_batches = (total + BATCH_SIZE - 1) // BATCH_SIZE
+    print(f"  -> {total} rows loaded")
+    print(f"  -> {BATCH_SIZE} rows per INSERT block")
+    print(f"  -> {num_batches} total batches\n")
+
+    header = f"""\
+-- Qdrant Query Language
+-- BULK INSERT -- DOCTOR PATIENT CONVERSATION (BATCHED)
+
+-- ============================================================
+--  QQL -- Doctor Patient Conversation
+--  Collection : {COLLECTION}
+--  Source     : {REPO_ID}
+--  Total rows : {total}
+--  Batch size : {BATCH_SIZE}
+--  Generated  : {TODAY}
+-- ============================================================
+
+-- Step 0: Show Collections
+SHOW COLLECTIONS
+
+-- Step 1: Create the collection
+CREATE COLLECTION {COLLECTION}
+
+-- ============================================================
+-- BULK INSERT -- BATCHED (each block <= {BATCH_SIZE} records)
+-- ============================================================
+"""
+
+    print(f"Writing {OUTPUT_FILE} ...")
+    with open(OUTPUT_FILE, "w", encoding="utf-8") as f:
+        f.write(header)
+
+        batch     = []
+        batch_num = 1
+
+        for i, row in enumerate(ds):
+            batch.append(build_record(row))
+
+            if len(batch) == BATCH_SIZE:
+                write_batch(f, batch, batch_num, COLLECTION)
+                print(f"  batch {batch_num} written ({i + 1}/{total} rows)")
+                batch     = []
+                batch_num += 1
+
+        # flush remaining rows
+        if batch:
+            write_batch(f, batch, batch_num, COLLECTION)
+            print(f"  batch {batch_num} written ({total}/{total} rows)")
+
+    print(f"\nDone. '{OUTPUT_FILE}' written -- {total} records across {batch_num} INSERT blocks.")
+
+
+if __name__ == "__main__":
+    main()

examples_and_applications/healthcare_conversation_rag/main.py

@@ -0,0 +1,46 @@
+from agno.agent import Agent
+from agno.models.ollama import Ollama
+from qql import Connection
+
+def search_medical_records(question: str) -> str:
+    """
+    Use this function when you need to retrieve relevant medical context
+    for a single natural language question from the 'medical_records'
+    collection stored in Qdrant via QQL.
+
+    This function performs a semantic similarity search against the
+    collection and aggregates the top matching document texts into a
+    single context string, which can then be passed to an LLM or used
+    for further processing (e.g. RAG pipelines, answer generation).
+
+    Args:
+        question (str): A single medical question in natural language.
+
+    Returns:
+        str: Aggregated context text from the top matching records.
+             Returns an empty string if no results are found or an
+             error occurs.
+    """
+    LIMIT = 5
+    CONTEXT = ""
+
+    try:
+        with Connection(url="http://localhost:6333", secret="th3s3cr3tk3y") as conn:
+            query = f"SEARCH medical_records SIMILAR TO '{question}' LIMIT {LIMIT} WITH {{ hnsw_ef: 128, mmr_diversity: 0.5, mmr_candidates: 50}}"
+            result = conn.run_query(query=query)
+            for hit in result.data:
+                CONTEXT += hit["payload"]["text"]
+    except Exception as e:
+        print(f"Search failed for question: '{question}' | Error: {e}")
+
+    return CONTEXT
+
+agent = Agent(
+    tools=[search_medical_records],
+    model=Ollama(id="qwen3.5:latest", host="http://localhost:11434", timeout=300),
+    markdown=True,
+    debug_mode=True,
+    reasoning=True,
+    enable_agentic_memory=True
+)
+agent.print_response("hi doctor I am just wondering what is abutting and abutment of the nerve root means in a back issue please explain what treatment is required for annular bulging and tear", stream=True)

examples_and_applications/healthcare_conversation_rag/pyproject.toml

@@ -0,0 +1,12 @@
+[project]
+name = "qql-application"
+version = "0.1.0"
+description = "QQL and its application"
+requires-python = ">=3.13"
+dependencies = [
+    "qql-cli>=2.4.1",
+    "datasets>=4.8.5",
+    "agno>=2.6.7",
+    "ollama>=0.6.2",
+    "openai>=2.37.0"
+]