RAG Pipeline

Monitor, analyze, and optimize your Retrieval Augmented Generation pipeline

Live

hr_policy_rag_pipeline

Active
Description
RAG pipeline for HR policies, handbooks, and compliance docs
Created At
May 10, 2025 09:15 AM
Last Updated
May 26, 2025 10:42 AM
Version
v2.3.1
Owner
Platform Admin

Pipeline Flow

1. Ingestion

Source Connectors

Documents
32,847
Chunks
156,231
Status
Completed
Last Run
10:35 AM

2m 14s

2. Chunking

Recursive Splitter

Chunk Size
800
Overlap
100
Chunks
156,231
Status
Completed
Last Run
10:35 AM

1m 02s

3. Embedding

text-embedding-3-large

Model
text-embedding-3-large
Dimensions
1536
Status
Completed
Last Run
10:36 AM

3m 18s

4. Vector Store

Pinecone (HR Policies)

Index
hr-policy-index
Vectors
156,231
Dimension
1536
Status
Completed
Last Run
10:36 AM

45s

5. Retrieval

Top-K Retriever

Top K
5
Score Threshold
0.75
Search Type
Hybrid
Status
Completed
Last Run
10:42 AM

110ms

6. Augmentation

Context Builder

Max Context Tokens
8K
Rerank Model
bge-reranker-large
Status
Completed
Last Run
10:42 AM

210ms

7. Generation

gpt-4o

Model
gpt-4o
Max Tokens
1024
Temperature
0.2
Status
Completed
Last Run
10:42 AM

1.35s

  • Ingestion
  • Processing
  • Storage
  • Retrieval
  • Augmentation
  • Generation
  • Completed
  • In Progress

Recent Executions

Query IDUserStatusLatencyDocumentsTokensTime
q_9f7a2b1cjane.doe@acme.comSuccess1.72s51,84210:42 AM
q_8c2d4e5fjohn.smith@acme.comSuccess1.96s41,75610:41 AM
q_1a2b3c4dhr.admin@acme.comSuccess1.63s62,12310:40 AM
q_5e6f7g8halice.wong@acme.comSuccess2.10s52,04510:39 AM
q_2h4j6k8lmike.brown@acme.comFailed0010:38 AM

Retrieval Quality (24h)

Avg. Retrieval Score

0.86High
  • Excellent (≥ 0.85)62%
  • Good (0.70 - 0.84)25%
  • Fair (0.50 - 0.69)10%
  • Poor (< 0.50)3%

Top Retrieved Documents (24h)

DocumentHit Count
Employee Handbook.pdf1,248
Leave Policy.docx987
Code of Conduct.pdf765
Remote Work Policy.pdf642
Benefits Guide.pdf531

Pipeline Details

Pipeline ID
pl_7f3a9b2e
Type
RAG Pipeline
Status
Active
Schedule
Every 15 minutes
Next Run
10:45 AM (in 2m 30s)
Environment
Production
Tags
HRPoliciesCompliance

Pipeline Performance (24h)

  • Success Rate

    98.7%

  • Avg. Latency

    1.84s

  • Avg. Tokens / Query

    1,942

  • Total Cost

    $18.42

Recent Alerts

  • High latency detected in Embedding step10:15 AM
  • Low retrieval score threshold hit frequently09:42 AM
  • No alerts

AI Recommendations

  • Increase chunk overlap to improve retrieval recall.

  • Consider updating reranker model for better accuracy.

  • Reduce max context tokens to optimize cost.

  • Pipeline performance is optimal.