Adapt MCP schemas for OpenAI

biomni/agent/a1.py

@@ -1879,19 +1879,38 @@ def wrapper(**kwargs) -> dict:
             # Map your types to Python types
             type_map = {"str": str, "int": int, "float": float, "bool": bool, "List[str]": list[str], "dict": dict}
 
+            def get_python_type(param_info):
+                """Convert Biomni schema type to Python type."""
+                param_type_str = param_info["type"]
+
+                # Handle array types with proper items inspection
+                if param_type_str == "array" and "items" in param_info:
+                    items_type = param_info["items"]["type"]
+                    if items_type == "string":
+                        return list[str]
+                    elif items_type == "number":
+                        return list[float]
+                    elif items_type == "integer":
+                        return list[int]
+                    elif items_type == "boolean":
+                        return list[bool]
+                    else:
+                        return list[str]  # default to string list
+
+                # Handle regular types
+                return type_map.get(param_type_str, str)
+
             # Add required parameters
             for param_info in required_params:
                 param_name = param_info["name"]
-                param_type_str = param_info["type"]
-                param_type = type_map.get(param_type_str, str)
+                param_type = get_python_type(param_info)
 
                 new_params.append(inspect.Parameter(param_name, inspect.Parameter.KEYWORD_ONLY, annotation=param_type))
 
             # Add optional parameters
             for param_info in optional_params:
                 param_name = param_info["name"]
-                param_type_str = param_info["type"]
-                param_type = type_map.get(param_type_str, str)
+                param_type = get_python_type(param_info)
 
                 # Make it optional
                 optional_type = param_type | None

biomni/llm.py

@@ -111,7 +111,7 @@ def get_llm(
             raise ImportError(  # noqa: B904
                 "langchain-openai package is required for Azure OpenAI models. Install with: pip install langchain-openai"
             )
-        API_VERSION = "2024-12-01-preview"
+        API_VERSION = "2025-04-01-preview"
         model = model.replace("azure-", "")
         return AzureChatOpenAI(
             openai_api_key=os.getenv("OPENAI_API_KEY"),

biomni/tool/bioengineering.py

@@ -231,7 +231,9 @@ def analyze_cell_migration_metrics(
     return log.strip()
 
 
-def perform_crispr_cas9_genome_editing(guide_rna_sequences, target_genomic_loci, cell_tissue_type):
+def perform_crispr_cas9_genome_editing(
+    guide_rna_sequences: list[str], target_genomic_loci: str, cell_tissue_type: str
+) -> str:
     """Simulates CRISPR-Cas9 genome editing process including guide RNA design, delivery, and analysis.
 
     Parameters

biomni/tool/cancer_biology.py

@@ -265,7 +265,7 @@ def analyze_ddr_network_in_cancer(expression_data_path, mutation_data_path, outp
     return research_log
 
 
-def analyze_cell_senescence_and_apoptosis(fcs_file_path):
+def analyze_cell_senescence_and_apoptosis(fcs_file_path: str) -> str:
     """Analyze flow cytometry data to quantify senescent and apoptotic cell populations.
 
     Parameters

biomni/tool/cell_biology.py

@@ -1,4 +1,4 @@
-def quantify_cell_cycle_phases_from_microscopy(image_paths, output_dir="./results"):
+def quantify_cell_cycle_phases_from_microscopy(image_paths: list[str], output_dir: str = "./results") -> str:
     """Quantify the percentage of cells in each cell cycle phase using Calcofluor white stained microscopy images.
 
     This function processes microscopy images where cell walls/septa are stained with Calcofluor white,
@@ -159,7 +159,9 @@ def quantify_cell_cycle_phases_from_microscopy(image_paths, output_dir="./result
     return log
 
 
-def quantify_and_cluster_cell_motility(image_sequence_path, output_dir="./results", num_clusters=3):
+def quantify_and_cluster_cell_motility(
+    image_sequence_path: str, output_dir: str = "./results", num_clusters: int = 3
+) -> str:
     """Quantify cell motility features from time-lapse microscopy images and cluster cells based on motility patterns.
 
     Parameters
@@ -571,7 +573,9 @@ def analyze_flow_cytometry_immunophenotyping(
     return log
 
 
-def analyze_mitochondrial_morphology_and_potential(morphology_image_path, potential_image_path, output_dir="./output"):
+def analyze_mitochondrial_morphology_and_potential(
+    morphology_image_path: str, potential_image_path: str, output_dir: str = "./output"
+) -> str:
     """Quantifies metrics of mitochondrial morphology and membrane potential from fluorescence microscopy images.
 
     Parameters

biomni/tool/genomics.py

@@ -798,7 +798,7 @@ def create_panhumanpy_env(env_name):
 # === Integration ===
 
 
-def create_scvi_embeddings_scRNA(adata_filename, batch_key, label_key, data_dir):
+def create_scvi_embeddings_scRNA(adata_filename: str, batch_key: str, label_key: str, data_dir: str) -> str:
     # Import scvi-tools correctly - the package name is still 'scvi' when installed
     try:
         import scvi
@@ -839,7 +839,7 @@ def create_scvi_embeddings_scRNA(adata_filename, batch_key, label_key, data_dir)
     return "\n".join(steps)
 
 
-def create_harmony_embeddings_scRNA(adata_filename, batch_key, data_dir):
+def create_harmony_embeddings_scRNA(adata_filename: str, batch_key: str, data_dir: str) -> str:
     # https://pypi.org/project/harmony-pytorch/
     from harmony import harmonize
 
@@ -911,7 +911,7 @@ def get_uce_embeddings_scRNA(
     return "\n".join(steps)
 
 
-def map_to_ima_interpret_scRNA(adata_filename, data_dir, custom_args=None):
+def map_to_ima_interpret_scRNA(adata_filename: str, data_dir: str, custom_args=None) -> str:
     """Map cell embeddings from the input dataset to the Integrated Megascale Atlas reference dataset using UCE embeddings."""
     from sklearn.neighbors import NearestNeighbors
 
@@ -1088,7 +1088,9 @@ def gene_set_enrichment_analysis(
         return f"An error occurred: {e}"
 
 
-def analyze_chromatin_interactions(hic_file_path, regulatory_elements_bed, output_dir="./output"):
+def analyze_chromatin_interactions(
+    hic_file_path: str, regulatory_elements_bed: str, output_dir: str = "./output"
+) -> str:
     """Analyze chromatin interactions from Hi-C data to identify enhancer-promoter interactions and TADs.
 
     Parameters
@@ -1421,7 +1423,9 @@ def get_matrix(chrom, balance_if_possible=True):
     return "\n".join(log)
 
 
-def analyze_comparative_genomics_and_haplotypes(sample_fasta_files, reference_genome_path, output_dir="./output"):
+def analyze_comparative_genomics_and_haplotypes(
+    sample_fasta_files: str, reference_genome_path: str, output_dir: str = "./output"
+) -> str:
     """Perform comparative genomics and haplotype analysis on multiple genome samples.
 
     This function aligns multiple genome samples to a reference, identifies variants,
@@ -1962,7 +1966,7 @@ def find_enriched_motifs_with_homer(
     return log
 
 
-def analyze_genomic_region_overlap(region_sets, output_prefix="overlap_analysis"):
+def analyze_genomic_region_overlap(region_sets: str, output_prefix: str = "overlap_analysis") -> str:
     """Analyze overlaps between two or more sets of genomic regions.
 
     Parameters

biomni/tool/tool_description/bioengineering.py

@@ -49,7 +49,8 @@
                 "nucleotides each) targeting the "
                 "genomic region of interest",
                 "name": "guide_rna_sequences",
-                "type": "List[str]",
+                "type": "array",
+                "items": {"type": "string"},
             },
             {
                 "default": None,
@@ -120,13 +121,15 @@
                 "default": None,
                 "description": "Time points at which drug concentrations were measured (in hours)",
                 "name": "time_points",
-                "type": "List[float] or numpy.ndarray",
+                "type": "array",
+                "items": {"type": "number"},
             },
             {
                 "default": None,
                 "description": "Measured drug concentration at each time point",
                 "name": "concentration_data",
-                "type": "List[float] or numpy.ndarray",
+                "type": "array",
+                "items": {"type": "number"},
             },
         ],
     },
@@ -192,13 +195,13 @@
                 "default": None,
                 "description": "Neural spiking activity data, shape (n_timepoints, n_neurons)",
                 "name": "neural_data",
-                "type": "numpy.ndarray",
+                "type": "array",
             },
             {
                 "default": None,
                 "description": "Behavioral data, shape (n_timepoints, n_behavioral_variables)",
                 "name": "behavioral_data",
-                "type": "numpy.ndarray",
+                "type": "array",
             },
         ],
     },
@@ -216,13 +219,14 @@
                 "simple example whole-cell model "
                 "will be used.",
                 "name": "ode_function",
-                "type": "callable",
+                "type": "object",
             },
             {
                 "default": "(0, 100)",
                 "description": "Tuple of (start_time, end_time) for the simulation.",
                 "name": "time_span",
-                "type": "tuple",
+                "type": "array",
+                "items": {"type": "number"},
             },
             {
                 "default": 1000,
@@ -247,7 +251,7 @@
                 "If array-like, order must match the "
                 "order expected by the ODE function.",
                 "name": "initial_conditions",
-                "type": "dict or array-like",
+                "type": "object",
             },
             {
                 "default": None,

biomni/tool/tool_description/cell_biology.py

@@ -16,7 +16,8 @@
                 "default": None,
                 "description": "List of file paths to microscopy images of cells stained with Calcofluor white",
                 "name": "image_paths",
-                "type": "List[str]",
+                "type": "array",
+                "items": {"type": "string"},
             }
         ],
     },

biomni/tool/tool_description/database.py

@@ -141,15 +141,17 @@
                 "default": None,
                 "description": "List of specific attributes to retrieve",
                 "name": "attributes",
-                "type": "List[str]",
+                "type": "array",
+                "items": {"type": "string"},
             },
         ],
         "required_parameters": [
             {
                 "default": None,
                 "description": "List of PDB identifiers to query",
                 "name": "identifiers",
-                "type": "List[str]",
+                "type": "array",
+                "items": {"type": "string"},
             }
         ],
     },

biomni/tool/tool_description/genomics.py

@@ -165,7 +165,8 @@
                 "default": None,
                 "description": "Custom command line arguments to pass to the UCE script",
                 "name": "custom_args",
-                "type": "List[str]",
+                "type": "array",
+                "items": {"type": "string"},
             },
         ],
         "required_parameters": [
@@ -329,7 +330,8 @@
                 "default": None,
                 "description": "Paths to FASTA files containing whole-genome sequences to be analyzed",
                 "name": "sample_fasta_files",
-                "type": "List[str]",
+                "type": "array",
+                "items": {"type": "string"},
             },
             {
                 "default": None,
@@ -650,7 +652,8 @@
                 "default": None,
                 "description": "List of Ensembl gene IDs (e.g., ['ENSG00000012048', 'ENSG00000012049'])",
                 "name": "ensembl_gene_ids",
-                "type": "List[str]",
+                "type": "array",
+                "items": {"type": "string"},
             }
         ],
     },

biomni/tool/tool_description/microbiology.py

@@ -247,7 +247,8 @@
                 "default": None,
                 "description": "Names of the biofilm samples corresponding to od_values",
                 "name": "sample_names",
-                "type": "List[str]",
+                "type": "array",
+                "items": {"type": "string"},
             },
             {
                 "default": 0,
@@ -270,7 +271,8 @@
                 "representing absorbance readings "
                 "for samples",
                 "name": "od_values",
-                "type": "List[float] or numpy.ndarray",
+                "type": "array",
+                "items": {"type": "number"},
             }
         ],
     },
@@ -427,25 +429,29 @@
                 "default": None,
                 "description": "Initial population sizes for each microbial species",
                 "name": "initial_populations",
-                "type": "List[int]",
+                "type": "array",
+                "items": {"type": "integer"},
             },
             {
                 "default": None,
                 "description": "Per capita growth rates for each species",
                 "name": "growth_rates",
-                "type": "List[float]",
+                "type": "array",
+                "items": {"type": "number"},
             },
             {
                 "default": None,
                 "description": "Per capita death/clearance rates for each species",
                 "name": "clearance_rates",
-                "type": "List[float]",
+                "type": "array",
+                "items": {"type": "number"},
             },
             {
                 "default": None,
                 "description": "Maximum sustainable population for each species",
                 "name": "carrying_capacities",
-                "type": "List[float]",
+                "type": "array",
+                "items": {"type": "number"},
             },
         ],
     },