Hugging Face's logo

Spaces:
19arjun89
/
AI_Recruiting_Agent
Running

App Files Community
1
19arjun89 commited on
Commit
1012d4d
·
verified ·
1 Parent(s): 5b3bd15

Update Guidebook.md

Browse files
Files changed (1) hide show
  1. Guidebook.md +24 -58
Guidebook.md CHANGED
@@ -1,60 +1,26 @@
1
- AI RECRUITING ASSISTANT — TABULAR PIPELINE (SWIM-LANE VIEW)
2
 
3
- +------+-------------------+----------------------------+------------------------------+------------------------------+
4
- | Step | Recruiter / Input | Python / Deterministic | LLM (Groq) | Storage / Output |
5
- +------+-------------------+----------------------------+------------------------------+------------------------------+
6
- | 1 | Upload culture | Chunk + embed | — | culture_store (indexed) |
7
- | | documents | | | |
8
- +------+-------------------+----------------------------+------------------------------+------------------------------+
9
- | 2 | Upload resumes | Anonymize → chunk → embed | — | resume_store (indexed) |
10
- | | | | | |
11
- +------+-------------------+----------------------------+------------------------------+------------------------------+
12
- | 3 | Paste JD + Run | Send JD to LLM | Extract required skills | required_skills JSON |
13
- | | | | + evidence quotes | |
14
- +------+-------------------+----------------------------+------------------------------+------------------------------+
15
- | 4 | — | Verify requirements | — | quote-verified list |
16
- | | | (quote/name/drop) | | + debug report |
17
- +------+-------------------+----------------------------+------------------------------+------------------------------+
18
- | 5 | — | Retrieve culture (k=3) | — | culture_context |
19
- | | | | | |
20
- +------+-------------------+----------------------------+------------------------------+------------------------------+
21
- | 6 | — | — | Generate culture attributes | cultural_attributes JSON |
22
- | | | | (job-performance aligned) | |
23
- +------+-------------------+----------------------------+------------------------------+------------------------------+
24
- | 7 | — | Retrieve resumes (k=10) | — | resume chunks by resume_id |
25
- | | | Group by candidate | | |
26
- +------+-------------------+----------------------------+------------------------------+------------------------------+
27
- | 8 | — | — | Propose culture matches | raw culture-match JSON |
28
- | | | | (direct/inferred + quotes) | |
29
- +------+-------------------+----------------------------+------------------------------+------------------------------+
30
- | 9 | — | Cleanup + weight scoring | — | culture_score + evidence |
31
- | | | (direct=1.0, inferred=0.5) | | |
32
- +------+-------------------+----------------------------+------------------------------+------------------------------+
33
- | 10 | — | — | Propose skill matches | raw skills-match JSON |
34
- | | | | + evidence snippets | |
35
- +------+-------------------+----------------------------+------------------------------+------------------------------+
36
- | 11 | — | Compute missing + score | — | skills_score + missing list |
37
- | | | (verified reqs only) | | |
38
- +------+-------------------+----------------------------+------------------------------+------------------------------+
39
- | 12 | — | — | Infer implied skills | implied (NOT SCORED) |
40
- | | | | + quotes + phone questions | follow-ups |
41
- +------+-------------------+----------------------------+------------------------------+------------------------------+
42
- | 13 | — | — | Verify evidence (✓/✗) | factuality report |
43
- | | | | | |
44
- +------+-------------------+----------------------------+------------------------------+------------------------------+
45
- | 14 | — | — | Generate recommendation | final recommendation |
46
- | | | | (policy constrained) | |
47
- +------+-------------------+----------------------------+------------------------------+------------------------------+
48
- | 15 | — | Trigger self-correction | Revise if needed | corrected recommendation |
49
- | | | (if unverified claims) | | + revision note |
50
- +------+-------------------+----------------------------+------------------------------+------------------------------+
51
- | 16 | — | — | Run bias audit | bias indicators + guidance |
52
- | | | | (JD + analyses + decision) | |
53
- +------+-------------------+----------------------------+------------------------------+------------------------------+
54
- | 17 | Review output | Assemble final report | — | full candidate report |
55
- | | | | | (scores, audits, evidence) |
56
- +------+-------------------+----------------------------+------------------------------+------------------------------+
57
 
58
- Current Retrieval Parameters:
59
- - Culture store: k = 3 chunks (JD query)
60
- - Resume store: k = 10 chunks (JD query)
 
 
1
+ ## End-to-End Pipeline (Swim-Lane View)
2
 
3
+ | Step | Recruiter / Input | Python / Deterministic Logic | LLM (Groq) | Storage / Output |
4
+ |------|------------------|------------------------------|-----------|------------------|
5
+ | 1 | Upload culture documents | Chunk + embed | — | `culture_store` (indexed) |
6
+ | 2 | Upload resumes | Anonymize → chunk → embed | — | `resume_store` (indexed) |
7
+ | 3 | Paste JD + Run | Send JD to LLM | Extract required skills + evidence quotes | `required_skills` JSON |
8
+ | 4 | — | Verify requirements (quote / name-only / dropped) | — | Verified list + debug report |
9
+ | 5 | — | Retrieve culture context (k=3) | — | `culture_context` |
10
+ | 6 | — | — | Generate culture attributes (job-performance aligned) | `cultural_attributes` JSON |
11
+ | 7 | — | Retrieve resume chunks (k=10), group by `resume_id` | — | Candidate chunks |
12
+ | 8 | — | — | Propose culture matches (direct/inferred + quotes) | Raw culture-match JSON |
13
+ | 9 | — | Cleanup + weighted scoring (direct=1.0, inferred=0.5) | — | Culture score + evidence |
14
+ | 10 | — | — | Propose skill matches + evidence snippets | Raw skills-match JSON |
15
+ | 11 | — | Compute missing list + skills score (verified reqs only) | — | Skills score + missing list |
16
+ | 12 | — | — | Infer implied skills (NOT SCORED) + phone questions | Implied follow-ups |
17
+ | 13 | — | — | Verify evidence (✓/✗) | Factuality report |
18
+ | 14 | — | — | Generate recommendation (policy constrained) | Final recommendation |
19
+ | 15 | — | Trigger self-correction (if needed) | Revise flagged claims only | Corrected recommendation |
20
+ | 16 | — | — | Run bias audit (JD + analyses + decision) | Bias indicators + guidance |
21
+ | 17 | Review output | Assemble final report | — | Full candidate report |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
22
 
23
+ ### Current Retrieval Parameters
24
+
25
+ - Culture store: `k = 3` chunks (JD query)
26
+ - Resume store: `k = 10` chunks (JD query)