feat: add structured filters to search_people

[arwinsoetanto]

Summary

• Add 6 new optional filter parameters to search_people: current_company, past_company, school, title, network, industry
• Add _build_people_search_url() static method following the existing _build_job_search_url() pattern
• Add _format_bracket_list() helper for LinkedIn's bracket-list URL syntax (e.g. ["103334640","162479"])
• Add _NETWORK_MAP for human-readable network degree values

Motivation

search_jobs has rich structured filters but search_people only has keywords + location. This makes people search unreliable for the primary MCP use case: LLM-driven LinkedIn research.

Real-world case study: I was using Claude Code with this MCP server to identify the hiring manager for a Product Manager role at a 74-person startup. Keyword searches like "Company Name" VP director head of product returned mostly people at other companies (fuzzy matching), and missed the actual founding Head of Product because she was a 3rd+ connection — LinkedIn's relevance ranking buried her behind 2nd-degree connections at unrelated companies.

A single structured query with current_company="103334640" would have returned all employees at that specific company, regardless of connection degree. This is exactly what LinkedIn's UI does when you click "Current company" in the filters panel.

LinkedIn URL parameter mapping

Parameter	URL param	Example
current_company	currentCompany	["103334640"]
past_company	pastCompany	["103334640"]
school	schoolFilter	["1790"]
title	titleFreeText	product%20manager
network	network	["F","S"]
industry	industry	["14"]

Implementation notes

• Follows the exact same pattern as _build_job_search_url() / search_jobs — filter maps, _normalize_csv(), URL parameter building
• _format_bracket_list() converts comma-separated IDs to LinkedIn's ["id1","id2"] bracket syntax
• _NETWORK_MAP maps human-readable "first", "second", "third" to LinkedIn's "F", "S", "O" codes
• All new parameters are optional — fully backward compatible, existing behavior unchanged when omitted
• Docstrings include guidance for LLMs on where to find company/school IDs (from prior get_company_profile or get_person_profile calls)
• Linter passes (ruff check)

Test plan

• _format_bracket_list("103334640,162479") produces ["103334640","162479"]
• _build_people_search_url(keywords="product", current_company="103334640", title="head", network="first,second") produces correct URL with encoded filters
• ruff check passes on both changed files
• Import verification succeeds
• Live test: search_people(keywords="", current_company="103334640") returns employees at the target company

🤖 Generated with Claude Code

[greptile-apps]

Greptile Summary

This PR adds 6 optional structured filter parameters (current_company, past_company, school, title, network, industry) to search_people, directly mirroring the established _build_job_search_url / search_jobs pattern. All new parameters are optional with None defaults, making the change fully backward compatible with existing callers.

Confidence Score: 5/5

Safe to merge — all changes are additive and backward compatible with no breaking changes to existing behavior.

The implementation closely follows the established _build_job_search_url pattern, all new parameters are optional with None defaults, and the only finding is a minor edge-case robustness issue in _format_bracket_list that is unlikely to affect real LLM usage.

No files require special attention.

Important Files Changed

Filename	Overview
linkedin_mcp_server/scraping/extractor.py	Adds _NETWORK_MAP, _format_bracket_list, and _build_people_search_url following the existing _build_job_search_url pattern; cleanly delegates to helper functions.
linkedin_mcp_server/tools/person.py	Adds 6 optional filter params to the search_people MCP tool and correctly forwards them to the extractor.

Sequence Diagram

sequenceDiagram
    participant LLM as LLM/Client
    participant Tool as search_people (person.py)
    participant Extractor as LinkedInExtractor
    participant Builder as _build_people_search_url
    participant Helpers as _format_bracket_list / _normalize_csv
    participant LinkedIn as LinkedIn Search

    LLM->>Tool: search_people(keywords, current_company, network, ...)
    Tool->>Extractor: extractor.search_people(keywords, current_company, network, ...)
    Extractor->>Builder: _build_people_search_url(keywords, current_company, network, ...)
    Builder->>Helpers: _format_bracket_list(current_company)
    Helpers-->>Builder: ["103334640"]
    Builder->>Helpers: _normalize_csv(network, _NETWORK_MAP) + _format_bracket_list
    Helpers-->>Builder: ["F","S"]
    Builder-->>Extractor: https://linkedin.com/search/results/people/?keywords=...
    Extractor->>LinkedIn: GET constructed URL
    LinkedIn-->>Extractor: search results page
    Extractor-->>Tool: {url, sections, references}
    Tool-->>LLM: {url, sections, references}

Loading

Prompt To Fix All With AI

This is a comment left during a code review.
Path: linkedin_mcp_server/scraping/extractor.py
Line: 153-155

Comment:
**Whitespace-only inputs produce malformed bracket entries**

If a caller passes a whitespace-only string (e.g. `"  "`), `if current_company:` is truthy, so `_format_bracket_list` is invoked and produces `[""]` — an invalid LinkedIn ID that would silently corrupt the filter. Filtering empty parts after stripping fixes this.

```suggestion
    parts = [v.strip() for v in value.split(",") if v.strip()]
    inner = ",".join('"' + p + '"' for p in parts)
    return "[" + inner + "]"
```

How can I resolve this? If you propose a fix, please make it concise.

_{Reviews (1): Last reviewed commit: "feat: add structured filters to search_p..." | Re-trigger Greptile}

[greptile-apps]

Whitespace-only inputs produce malformed bracket entries

If a caller passes a whitespace-only string (e.g. " "), if current_company: is truthy, so _format_bracket_list is invoked and produces [""] — an invalid LinkedIn ID that would silently corrupt the filter. Filtering empty parts after stripping fixes this.

Suggested change

	parts = [v.strip() for v in value.split(",")]
	inner = ",".join('"' + p + '"' for p in parts)
	return "[" + inner + "]"
	parts = [v.strip() for v in value.split(",") if v.strip()]
	inner = ",".join('"' + p + '"' for p in parts)
	return "[" + inner + "]"

Prompt To Fix With AI

This is a comment left during a code review.
Path: linkedin_mcp_server/scraping/extractor.py
Line: 153-155

Comment:
**Whitespace-only inputs produce malformed bracket entries**

If a caller passes a whitespace-only string (e.g. `"  "`), `if current_company:` is truthy, so `_format_bracket_list` is invoked and produces `[""]` — an invalid LinkedIn ID that would silently corrupt the filter. Filtering empty parts after stripping fixes this.

```suggestion
    parts = [v.strip() for v in value.split(",") if v.strip()]
    inner = ",".join('"' + p + '"' for p in parts)
    return "[" + inner + "]"
```

How can I resolve this? If you propose a fix, please make it concise.