All templates By NappAI team CV Match
A flow that compares CVs with requirements, calculates matches, and returns the candidate with the best fit.
+4
graph TD
%%{init: {'theme': 'mc','layout': 'elk'}}%%
TextInput-dvpjv[<div><img src="/_astro/type.Dy26vmDy.svg" style="height: 20px !important;width: 20px !important"/></div>Carrera]
style TextInput-dvpjv stroke:#a170ff
TextInput-s2vgd[<div><img src="/_astro/type.Dy26vmDy.svg" style="height: 20px !important;width: 20px !important"/></div>Certificados]
style TextInput-s2vgd stroke:#a170ff
TextInput-oz3io[<div><img src="/_astro/type.Dy26vmDy.svg" style="height: 20px !important;width: 20px !important"/></div>Habilidades]
style TextInput-oz3io stroke:#a170ff
TextInput-6r3sb[<div><img src="/_astro/type.Dy26vmDy.svg" style="height: 20px !important;width: 20px !important"/></div>Experiencia]
style TextInput-6r3sb stroke:#a170ff
TextInput-ep0wz[<div><img src="/_astro/type.Dy26vmDy.svg" style="height: 20px !important;width: 20px !important"/></div>Educación]
style TextInput-ep0wz stroke:#a170ff
OpenAIModel-a5e3b[<div><img src="/_astro/openAI.BhmuxEs3.svg" style="height: 20px !important;width: 20px !important"/></div>OpenAI]
style OpenAIModel-a5e3b stroke:#a170ff
ChatOutput-zlpn1[<div><img src="/_astro/messages-square.BaSDmT6g.svg" style="height: 20px !important;width: 20px !important"/></div>Chat Output]
style ChatOutput-zlpn1 stroke:#a170ff
ParseData-4y52m[<div><img src="/_astro/braces.Djq0PW4_.svg" style="height: 20px !important;width: 20px !important"/></div>Obtener Texto]
style ParseData-4y52m stroke:#a170ff
OpenAIEmbeddings-5myel[<div><img src="/_astro/openAI.BhmuxEs3.svg" style="height: 20px !important;width: 20px !important"/></div>OpenAI Embeddings]
style OpenAIEmbeddings-5myel stroke:#a170ff
GDriveFilesComponent-3xa20[<div><img src="/_astro/google_drive.wKmDsV2c.svg" style="height: 20px !important;width: 20px !important"/></div>Obtener CV]
style GDriveFilesComponent-3xa20 stroke:#a170ff
Prompt-3a8sl[<div><img src="/_astro/square-terminal.BMOXc-nZ.svg" style="height: 20px !important;width: 20px !important"/></div>Requisitos]
style Prompt-3a8sl stroke:#a170ff
Prompt-p6nvb[<div><img src="/_astro/square-terminal.BMOXc-nZ.svg" style="height: 20px !important;width: 20px !important"/></div>Instrucciones]
style Prompt-p6nvb stroke:#a170ff
Chroma-regly[<div><img src="/_astro/chroma.CDTUBZSx.svg" style="height: 20px !important;width: 20px !important"/></div>Subir a DB]
style Chroma-regly stroke:#a170ff
OpenAIEmbeddings-6p7a2[<div><img src="/_astro/openAI.BhmuxEs3.svg" style="height: 20px !important;width: 20px !important"/></div>OpenAI Embeddings2]
style OpenAIEmbeddings-6p7a2 stroke:#a170ff
Chroma-r9ojf[<div><img src="/_astro/chroma.CDTUBZSx.svg" style="height: 20px !important;width: 20px !important"/></div>Cargar de DB]
style Chroma-r9ojf stroke:#a170ff
LanguageRecursiveTextSplitter-rweqg[Language Recursive Text Splitter]
style LanguageRecursiveTextSplitter-rweqg stroke:#a170ff
OpenAIModel-a5e3b -.- ChatOutput-zlpn1
linkStyle 0 stroke:#a170ff
TextInput-dvpjv -.- Prompt-3a8sl
linkStyle 1 stroke:#a170ff
TextInput-s2vgd -.- Prompt-3a8sl
linkStyle 2 stroke:#a170ff
TextInput-ep0wz -.- Prompt-3a8sl
linkStyle 3 stroke:#a170ff
TextInput-6r3sb -.- Prompt-3a8sl
linkStyle 4 stroke:#a170ff
TextInput-oz3io -.- Prompt-3a8sl
linkStyle 5 stroke:#a170ff
Prompt-3a8sl -.- Prompt-p6nvb
linkStyle 6 stroke:#a170ff
ParseData-4y52m -.- Prompt-p6nvb
linkStyle 7 stroke:#a170ff
Prompt-p6nvb -.- OpenAIModel-a5e3b
linkStyle 8 stroke:#a170ff
OpenAIEmbeddings-5myel -.- Chroma-regly
linkStyle 9 stroke:#a170ff
OpenAIEmbeddings-6p7a2 -.- Chroma-r9ojf
linkStyle 10 stroke:#a170ff
Chroma-r9ojf -.- ParseData-4y52m
linkStyle 11 stroke:#a170ff
GDriveFilesComponent-3xa20 -.- LanguageRecursiveTextSplitter-rweqg
linkStyle 12 stroke:#a170ff
LanguageRecursiveTextSplitter-rweqg -.- Chroma-regly
linkStyle 13 stroke:#a170ff
🧩 Overview
The workflow automates the comparison of candidate CVs against a set of user‑defined requirements. It retrieves CV documents from Google Drive, splits them into manageable chunks, embeds the text, stores the embeddings in a Chroma vector database, retrieves the most relevant CVs, extracts structured data, constructs prompts from user inputs, and finally generates a recommendation for the best‑matching candidate using an OpenAI language model.
⚙️ Main Features
- Retrieves and ingests CV files from a specified Google Drive folder.
- Splits CV text into language‑aware chunks for efficient embedding.
- Generates embeddings for both ingestion and search using OpenAI embeddings.
- Stores embeddings in a Chroma vector store and retrieves relevant CVs.
- Parses retrieved CV data into a plain‑text format.
- Builds dynamic prompts from user inputs (education, experience, etc.).
- Generates matching instructions with an OpenAI language model.
- Displays the final recommendation in a chat output.
🔄 Workflow Steps
| Component Name | Role in the Workflow | Key Inputs | Key Outputs |
|---|---|---|---|
| Obtener CV | Reads CV files from a Google Drive folder and delivers them to the pipeline. | Google Drive folder path, file selection settings. | CV file content as Data. |
| Language Recursive Text Splitter | Splits the CV text into language‑aware chunks for embedding. | CV Data. | Chunked Data. |
| OpenAI Embeddings (ingest) | Creates embeddings for each chunk to be stored in the vector database. | Chunked Data. | Embedding vectors. |
| Subir a DB | Ingests chunked data and their embeddings into a Chroma vector store. | Chunked Data, Embedding vectors. | Stored embeddings in Chroma. |
| OpenAI Embeddings (search) | Generates embeddings for the search query (CV). | Search query Data (CV). | Embedding vectors for search. |
| Cargar de DB | Retrieves the most similar CVs from Chroma based on the search embeddings. | Search query Data, embedding vectors. | Search results as Data. |
| ParseData | Converts the retrieved CV data into plain text suitable for prompt construction. | Retrieved CV Data. | Plain‑text Data (CV text). |
| Carrera (Text Input) | Captures the candidate’s field of study. | User input. | Text. |
| Certificados (Text Input) | Captures the candidate’s certifications. | User input. | Text. |
| Educación (Text Input) | Captures the candidate’s education details. | User input. | Text. |
| Experiencia (Text Input) | Captures the candidate’s work experience. | User input. | Text. |
| Habilidades (Text Input) | Captures the candidate’s skills. | User input. | Text. |
| Requisitos | Builds a prompt template containing placeholders for the user inputs. | Carrera, Certificados, Educación, Experiencia, Habilidades. | Prompt message with placeholders. |
| Instrucciones | Generates a user‑friendly prompt that includes the CV text and the user’s input. | CV text (from ParseData), user’s query (from Requisitos). | Final prompt message. |
| OpenAI Model | Generates the recommendation by responding to the prompt. | Prompt message. | Generated text. |
| Chat Output | Displays the recommendation in the playground chat. | Generated text. | Chat message shown to the user. |
All components are executed sequentially as dictated by the directed edges in the workflow diagram.
🧠 Notes
- Credentials: The workflow requires valid Google Drive and OpenAI API credentials.
- Embeddings: Two separate OpenAI embeddings components are used; one for ingesting CV chunks and one for generating search queries.
- Vector Store: Chroma is configured to persist embeddings in the
CV_matchdirectory and to retrieve up to 10 similar documents. - Chunking: The splitter uses a chunk size of 1000 tokens with 200 token overlap to preserve context.
- Prompt Generation: The
Requisitoscomponent dynamically inserts user inputs into a template that describes the required CV attributes. - Error Handling: If no matching CVs are found, the ParseData component will return an empty text, causing the OpenAI model to respond accordingly.
- Performance: Embedding generation and database ingestion are the most compute‑heavy steps; caching is enabled for these components to reduce repeated calls.
This documentation provides a concise, functional overview of the workflow, its components, and their interactions without delving into internal implementation details.