Update mcp/orchestrator.py
Browse files- mcp/orchestrator.py +24 -22
mcp/orchestrator.py
CHANGED
@@ -21,7 +21,7 @@ from mcp.embeddings import embed_texts, cluster_embeddings
|
|
21 |
|
22 |
def _get_llm(llm: str):
|
23 |
"""
|
24 |
-
|
25 |
"""
|
26 |
if llm.lower() == "gemini":
|
27 |
return gemini_summarize, gemini_qa
|
@@ -30,27 +30,27 @@ def _get_llm(llm: str):
|
|
30 |
|
31 |
async def orchestrate_search(query: str, llm: str = "openai") -> Dict[str, Any]:
|
32 |
"""
|
33 |
-
|
34 |
-
embeddings
|
35 |
"""
|
36 |
-
#
|
37 |
arxiv_task = fetch_arxiv(query)
|
38 |
pubmed_task = fetch_pubmed(query)
|
39 |
-
|
40 |
papers: List[Dict] = []
|
41 |
-
for res in
|
42 |
if isinstance(res, list):
|
43 |
papers.extend(res)
|
44 |
|
45 |
-
#
|
46 |
blob = " ".join(p.get("summary", "") for p in papers)
|
47 |
umls = await extract_umls_concepts(blob)
|
48 |
|
49 |
-
#
|
50 |
-
|
51 |
-
umls_relations = await asyncio.gather(*
|
52 |
|
53 |
-
#
|
54 |
names = [c["name"] for c in umls]
|
55 |
fda_tasks = [fetch_drug_safety(n) for n in names]
|
56 |
gene_task = search_gene(names[0]) if names else asyncio.sleep(0, result=[])
|
@@ -60,6 +60,7 @@ async def orchestrate_search(query: str, llm: str = "openai") -> Dict[str, Any]:
|
|
60 |
ot_task = ot.fetch(names[0]) if names else asyncio.sleep(0, result=[])
|
61 |
cbio_task = cbio.fetch_variants(names[0]) if names else asyncio.sleep(0, result=[])
|
62 |
|
|
|
63 |
fda, gene, mesh, dis, trials, ot_assoc, variants = await asyncio.gather(
|
64 |
asyncio.gather(*fda_tasks, return_exceptions=True),
|
65 |
gene_task, mesh_task, dis_task,
|
@@ -67,16 +68,17 @@ async def orchestrate_search(query: str, llm: str = "openai") -> Dict[str, Any]:
|
|
67 |
return_exceptions=False
|
68 |
)
|
69 |
|
70 |
-
#
|
71 |
summaries = [p.get("summary", "") for p in papers]
|
72 |
if summaries:
|
73 |
-
|
74 |
-
clusters = await cluster_embeddings(
|
|
|
|
|
75 |
else:
|
76 |
-
|
77 |
-
clusters = []
|
78 |
|
79 |
-
#
|
80 |
summarize_fn, _ = _get_llm(llm)
|
81 |
try:
|
82 |
ai_summary = await summarize_fn(blob)
|
@@ -94,7 +96,7 @@ async def orchestrate_search(query: str, llm: str = "openai") -> Dict[str, Any]:
|
|
94 |
"clinical_trials": trials,
|
95 |
"ot_associations": ot_assoc,
|
96 |
"variants": variants,
|
97 |
-
"embeddings":
|
98 |
"clusters": clusters,
|
99 |
"ai_summary": ai_summary,
|
100 |
"llm_used": llm.lower()
|
@@ -103,11 +105,11 @@ async def orchestrate_search(query: str, llm: str = "openai") -> Dict[str, Any]:
|
|
103 |
|
104 |
async def answer_ai_question(question: str, context: str = "", llm: str = "openai") -> Dict[str, str]:
|
105 |
"""
|
106 |
-
Follow-up Q&A
|
107 |
"""
|
108 |
_, qa_fn = _get_llm(llm)
|
109 |
try:
|
110 |
-
|
111 |
except Exception:
|
112 |
-
|
113 |
-
return {"answer":
|
|
|
21 |
|
22 |
def _get_llm(llm: str):
|
23 |
"""
|
24 |
+
Route summarization and QA to the chosen engine.
|
25 |
"""
|
26 |
if llm.lower() == "gemini":
|
27 |
return gemini_summarize, gemini_qa
|
|
|
30 |
|
31 |
async def orchestrate_search(query: str, llm: str = "openai") -> Dict[str, Any]:
|
32 |
"""
|
33 |
+
Fetch papers, extract concepts & relations, enrich data,
|
34 |
+
compute embeddings+clusters, and run LLM summary.
|
35 |
"""
|
36 |
+
# Gather literature
|
37 |
arxiv_task = fetch_arxiv(query)
|
38 |
pubmed_task = fetch_pubmed(query)
|
39 |
+
lit_results = await asyncio.gather(arxiv_task, pubmed_task, return_exceptions=True)
|
40 |
papers: List[Dict] = []
|
41 |
+
for res in lit_results:
|
42 |
if isinstance(res, list):
|
43 |
papers.extend(res)
|
44 |
|
45 |
+
# Concept extraction
|
46 |
blob = " ".join(p.get("summary", "") for p in papers)
|
47 |
umls = await extract_umls_concepts(blob)
|
48 |
|
49 |
+
# Fetch UMLS relations
|
50 |
+
rel_tasks = [fetch_relations(c["cui"]) for c in umls]
|
51 |
+
umls_relations = await asyncio.gather(*rel_tasks, return_exceptions=True)
|
52 |
|
53 |
+
# Data enrichment tasks
|
54 |
names = [c["name"] for c in umls]
|
55 |
fda_tasks = [fetch_drug_safety(n) for n in names]
|
56 |
gene_task = search_gene(names[0]) if names else asyncio.sleep(0, result=[])
|
|
|
60 |
ot_task = ot.fetch(names[0]) if names else asyncio.sleep(0, result=[])
|
61 |
cbio_task = cbio.fetch_variants(names[0]) if names else asyncio.sleep(0, result=[])
|
62 |
|
63 |
+
# Run enrichment
|
64 |
fda, gene, mesh, dis, trials, ot_assoc, variants = await asyncio.gather(
|
65 |
asyncio.gather(*fda_tasks, return_exceptions=True),
|
66 |
gene_task, mesh_task, dis_task,
|
|
|
68 |
return_exceptions=False
|
69 |
)
|
70 |
|
71 |
+
# Embeddings & clustering
|
72 |
summaries = [p.get("summary", "") for p in papers]
|
73 |
if summaries:
|
74 |
+
embeddings = await embed_texts(summaries)
|
75 |
+
clusters = await cluster_embeddings(
|
76 |
+
embeddings, n_clusters = max(2, min(10, len(embeddings)//2))
|
77 |
+
)
|
78 |
else:
|
79 |
+
embeddings, clusters = [], []
|
|
|
80 |
|
81 |
+
# LLM summary
|
82 |
summarize_fn, _ = _get_llm(llm)
|
83 |
try:
|
84 |
ai_summary = await summarize_fn(blob)
|
|
|
96 |
"clinical_trials": trials,
|
97 |
"ot_associations": ot_assoc,
|
98 |
"variants": variants,
|
99 |
+
"embeddings": embeddings,
|
100 |
"clusters": clusters,
|
101 |
"ai_summary": ai_summary,
|
102 |
"llm_used": llm.lower()
|
|
|
105 |
|
106 |
async def answer_ai_question(question: str, context: str = "", llm: str = "openai") -> Dict[str, str]:
|
107 |
"""
|
108 |
+
Follow-up Q&A via chosen LLM.
|
109 |
"""
|
110 |
_, qa_fn = _get_llm(llm)
|
111 |
try:
|
112 |
+
ans = await qa_fn(question, context)
|
113 |
except Exception:
|
114 |
+
ans = "LLM follow-up failed."
|
115 |
+
return {"answer": ans}
|