Drug Pipeline Search Skill

This skill converts natural language questions into structured API queries against a pharmaceutical drug database, then presents the results in a readable format.

Workflow

1. Parse user intent — Extract key entities from the user's question
2. Build query parameters — Map entities to the query schema below
3. Execute the query — Run

   scripts/search.py

4. Present results — Format and display drug records to the user

Step 1: Extract Keywords

Identify the following entity types from the user's question:

drug_name

dict

{"logic": "or", "data": ["pembrolizumab"]}

company

List[str]

["Pfizer", "Roche"]

indication

List[str]

["lung cancer", "NSCLC"]

target

dict

{"logic": "or", "data": ["PD-1", "VEGF"]}

drug_modality

dict

{"logic": "or", "data": ["antibody", "small molecule"]}

drug_feature

dict

{"logic": "or", "data": ["bispecific"]}

phase

List[str]

["Preclinical", "I", "II", "III", "IV", "Others", "IND", "Suspended", "Approved", "Unknow", "Withdraw from Market", "BLA/NDA"]

location

List[str]

["China", "United States", "Japan"]

route_of_administration

dict

{"logic": "or", "data": ["IV", "oral"]}

page_num

int

0

page_size

int

200

Dict field format:

{"logic": "or", "data": ["value1", "value2"]}

• logic

  controls how multiple values are combined:

  "or"

  (any match) or

  "and"

  (all must match). Default to

  "or"

  unless the user explicitly wants all terms to apply simultaneously.

• data

  is the list of keyword strings to match.

Type rules:

• company

  ,

  indication

  ,

  phase

  ,

  location

  → plain

  List[str]

• drug_name

  ,

  target

  ,

  drug_modality

  ,

  drug_feature

  ,

  route_of_administration

  →

  dict

  with

  logic

  and

  data

• Default to

  page_num: 0, page_size: 10

  unless the user specifies otherwise
• Prefer English keywords (the database is indexed in English); translate non-English terms

Step 2: Execute the Query

python scripts/search.py --params '<JSON string>'

Or using a parameter file:

python scripts/search.py --params-file /tmp/query.json

Add

--raw

Step 3: Interpret Results

The response contains:

• page_num

  /

  page_size

  — current pagination state

• results

  — current page of drug records, each with name, phase, modality, targets, companies, indication, development progress, etc.

Step 4: Review and Fallback Search Strategies

If no results are returned, apply the fallback strategies below before giving up. When an initial query returns zero or poor results, try these strategies in order:

Strategy 1 — Drug Name Variant Expansion

Drug names in the database may use different formats (with/without hyphens, partial codes, aliases). Expand the

drug_name

{
  "drug_name": {"logic": "or", "data": ["SHR-A1904", "SHR A1904", "A1904", "SHR1904"]},
  "page_num": 0,
  "page_size": 50
}

Common variant patterns to try:

• Remove or replace hyphens:

  SHR-A1904

  →

  SHR A1904

  ,

  SHRA1904

• Strip prefix/suffix:

  9MW-2821

  →

  MW-2821

  ,

  9MW2821

• Known alias: include trade names or INN alongside internal codes

Strategy 2 — Company-First with Application-Layer Filtering

When drug name matching is unreliable, use the company as the anchor. Fetch a broad set of the company's drugs, then filter by modality/indication/target in post-processing.

{
  "company": ["Roche", "Roche Inc"],
  "page_num": 0,
  "page_size": 500
}

After retrieving results, apply local filters:

• modality == "Antibody-Drug Conjugates, ADCs"

• indication contains "breast cancer"

• drug_name matches known code pattern

Use this strategy when the drug code is ambiguous or the API match rate is low.

Strategy 3 — Broad Target/Modality Search with Post-Filtering

When neither name nor company is reliable, search by biological target and modality, then narrow results client-side.

{
  "target": {"logic": "or", "data": ["CLDN18.2", "Nectin-4", "HER2"]},
  "drug_modality": {"logic": "or", "data": ["Antibody-Drug Conjugates, ADCs"]},
  "page_num": 0,
  "page_size": 200
}

After retrieval, filter by company name or drug code pattern using substring matching (e.g. code starts with

SHR

9MW

A166

Note: If the API supports regex, patterns like

(SHR|9MW|A166)

can be passed directly in

drug_name.data

to broaden matching in a single call.

Decision Tree

Initial query returns results?
├── Yes → present results
└── No  → Strategy 1: expand drug_name variants
          └── Still no results → Strategy 2: company anchor + local filter
                                 └── Still no results → Strategy 3: target/modality broad search
Any step hits HTTP 429?
└── Pause entire chain 30s → resume from current strategy
    (sleep ≥5s between every request to avoid triggering 429)

Conversion Examples

User: "Find PD-1 antibodies in Phase 3"

{
  "target": {"logic": "or", "data": ["PD-1"]},
  "drug_modality": {"logic": "or", "data": ["antibody"]},
  "phase": ["III"],
  "page_num": 0,
  "page_size": 30
}

User: "Roche bispecific antibodies for lung cancer"

{
  "company": ["Roche"],
  "drug_feature": {"logic": "or", "data": ["bispecific"]},
  "indication": ["lung cancer"],
  "page_num": 0,
  "page_size": 30
}

User: "Oral small molecule KRAS G12C inhibitors"

{
  "target": {"logic": "or", "data": ["KRAS G12C"]},
  "drug_modality": {"logic": "or", "data": ["small molecule"]},
  "route_of_administration": {"logic": "or", "data": ["oral"]},
  "page_num": 0,
  "page_size": 30
}

User: "Drugs targeting both PD-1 and VEGF"

{
  "target": {"logic": "and", "data": ["PD-1", "VEGF"]},
  "page_num": 0,
  "page_size": 30
}

User: "Look up pembrolizumab"

{
  "drug_name": {"logic": "or", "data": ["pembrolizumab"]},
  "page_num": 0,
  "page_size": 30
}

Dependencies

• Python 3.8+

• requests

  library (

  pip install requests

  )
• Environment variable

  NOAH_API_TOKEN

  — API authentication token (required)
  • Register for a free account at noah.bio to obtain your API key.

Security & Packaging Notes

• This skill only calls NoahAI official HTTPS endpoints under

  https://www.noah.bio/api/

  and does not contact third-party services.
• It requires exactly one environment variable:

  NOAH_API_TOKEN

  . Store it in the environment or a local

  .env

  file, and never place it inline in commands, chats, or packaged files.
• The token is scoped to read medical public details only and cannot access private user records.
• The skill does not intentionally persist request parameters locally. Any server-side retention is determined by the NoahAI API service and its operational logging policies.
• It does not request persistent or system-level privileges and does not modify system configuration.
• The skill is source-file based (Python scripts only) and does not require runtime installs, package downloads, or external bootstrap steps.