Skip to content

Lexi

Lexi API on this instance: Swagger · ReDoc · Live OpenAPI JSON

Lexi HTTP reference

API version

Lexi API: 1.13.3

Lexi uses AI Hub to transform HR information into actionable skills data: it generates skills for jobs, assigns skills to jobs, normalizes labels and detects duplicates.

This section intentionally focuses on functional behavior. It explains what each endpoint does and the main processing stages without delving into code details.

Access

Lexi endpoints are exposed under:

/lexi/api/v1

The technical OpenAPI documentation is available at:

/lexi/docs

All Lexi endpoints require an API key.

Endpoints

Endpoint Functional purpose
POST /jobs/generate_skills Generate skills for a list of jobs, then enrich the result.
POST /jobs/get_skill_levels Generate only the expected levels for skills already assigned to jobs.
POST /jobs/get_skill_descriptions Generate only descriptions for a list of skills.
POST /jobs/get_skill_level_descriptions Generate contextual descriptions for four or five proficiency levels.
POST /jobs/suggest_missions Suggest missing missions for jobs in an existing catalog.
POST /jobs/suggest_skills Suggest new competencies for an existing job catalog, optionally focused on a theme.
POST /jobs/assign_families Classify jobs into a supplied list of job families.
POST /skills/assign_to_jobs Assign skills to the most relevant jobs.
POST /skills/normalize_labels Convert skill labels into clean, consistent labels.
POST /find_duplicates Group labels or objects that express the same meaning.

Generation workflow overview

The main endpoint is generate_skills. It operates as a processing pipeline:

  1. Lexi sends the jobs to process.
  2. AI Hub prepares the generation rules: language, groups, implicit types, number of skills per job and client context.
  3. A first LLM creates a global skill catalog for all jobs.
  4. If the catalog is too small, a second call adds only the missing skills.
  5. The catalog is then assigned job by job.
  6. Another LLM checks that the assigned skills are relevant.
  7. Definitions and levels are generated concurrently.
  8. The final result is sent back to Lexi through a callback.