Sold by: Unstract
Deployed on AWS
LLMWhisperer turns PDFs, scanned documents, images and spreadsheets into layout-preserving text that large language models can accurately read. Our powerful OCR supports multiple file formats, languages, and advanced capabilities like highlighting and webhooks.
This appliance runs the complete LLMWhisperer stack on a single EC2 instance in your own AWS account: documents are processed on the instance and are not sent to Unstract. Billing is per page processed, based on the mode used for extraction.
Overview
LLMWhisperer prepares complex documents for large language models. It extracts text while keeping the document's layout, so columns, tables, forms and checkboxes stay readable when passed to an LLM instead of collapsing into an unordered stream of words.
This appliance runs the complete LLMWhisperer stack on one EC2 instance in your AWS account. The API, the processing workers, the OCR engines and the database all run on the instance, and your documents never leave it.
Extraction modes (each billed per page):
- native_text: fast extraction from PDFs that already have a text layer.
- low_cost: OCR for scanned documents and images at the lowest cost.
- high_quality: higher-accuracy OCR for scanned, photographed or handwritten documents.
- form: extraction that preserves forms and checkboxes. Excel files (XLS/XLSX) are also processed and billed under this mode.
- table: extraction tuned for table-heavy documents.
How it works:
- On first boot the instance generates its own credentials, obtains a license tied to your AWS Marketplace subscription, and starts the stack. No manual activation is needed.
- You call the LLMWhisperer REST API (v2) on port 3006, authenticated with an API key unique to the instance.
- A web dashboard on port 3007 shows the license status.
- Pages processed are reported to AWS Marketplace every hour and appear on your AWS bill.
The first high_quality, form or table request after startup can take several minutes while OCR engines initialise.
Instance sizing: On first boot, the appliance sizes its processing workers and OCR engine replicas to the instance CPU and memory. Larger machines can process more pages in parallel, within worker and memory limits. This is vertical scaling within one EC2 instance, not automatic instance resizing or multi-instance scaling. To change size, launch a new instance; an existing instance keeps its first-boot configuration.
Billing (per page, by extraction mode):
- Each PDF or image page processed counts as one page.
- Excel files (XLS/XLSX) are billed in form mode. Each sheet counts as one page per 3,000 words of content, and at least one page.
- There is no monthly cap, minimum or commitment. The appliance applies no API rate limit: throughput depends on the instance size you choose.
- EC2 instance charges are billed separately by AWS.
Launch options:
- CloudFormation: a template creates the instance, an IAM instance role that allows aws-marketplace:MeterUsage, and a security group opening the API only to the range you enter; dashboard access uses a separate optional trusted CIDR and is closed by default.
- AMI: launch from the EC2 console and attach an IAM instance role that allows aws-marketplace:MeterUsage. One-click launch does not attach a role, and without it the appliance cannot obtain its license and will not start.
Requirements: outbound HTTPS from the instance to the Unstract license service and to AWS Marketplace metering. See the usage instructions for details.
Deployment, sizing and startup status: LLMWhisperer on AWS Marketplace
Logs and common issues: AWS Marketplace troubleshooting
Highlights
- Layout-preserving text extraction from PDFs, scans, images and Excel files, ready for LLMs and RAG pipelines.
- Runs entirely in your AWS account on a single EC2 instance; documents are processed on the instance and never sent to Unstract.
- Pay per page processed, per extraction mode, billed through your AWS invoice. Licensing is automatic from your Marketplace subscription.
Details
Sold by
Delivery method
Delivery option
LLMWhisperer appliance (CloudFormation)
64-bit (x86) Amazon Machine Image (AMI)
Latest version
Operating system
AmazonLinux 2023
Deployed on AWS
New
Introducing multi-product solutions
You can now purchase comprehensive solutions tailored to use cases and industries.
Features and programs
Financing for AWS Marketplace purchases
AWS Marketplace now accepts line of credit payments through the PNC Vendor Finance program. This program is available to select AWS customers in the US, excluding NV, NC, ND, TN, & VT.
Pricing
Pricing is based on actual usage, with charges varying according to how much you consume. Subscriptions have no end date and may be canceled any time.
Additional AWS infrastructure costs may apply. Use the AWS Pricing Calculator to estimate your infrastructure costs.
Dimension | Description | Cost/unit |
|---|---|---|
native_text | Pages processed in native_text mode | $0.001 |
low_cost | Pages processed in low_cost mode | $0.005 |
high_quality | Pages processed in high_quality mode | $0.01 |
form | Pages processed in form mode, including Excel (XLS/XLSX) files | $0.015 |
table | Pages processed in table mode | $0.015 |
Dimensions summary
You pay per page processed, and you pick a mode based on your document type. Each dimension is a separate processing mode, not a tier you commit to. Native_text handles digital PDFs with embedded text. Low_cost handles clean printed scans. High_quality handles low-quality scans, handwriting, and photos. Form processes checkboxes, radio buttons, and Excel files. Table processes structured tables. You can choose a different mode per request, so pricing scales with how many pages you run through each mode. There is no monthly cap.
Top-of-mind questions for buyers
You are charged per page processed in your chosen mode. For Excel files in form mode, each XLS or XLSX sheet counts as a page. You are only charged for pages that process successfully. Pages that fail from a technical error on the service side are not charged.
Each mode bills independently per page. You can pick a different mode per request, so pages run through native_text, low_cost, high_quality, form, and table are each metered at that mode's rate. Your total is the sum across all modes. Higher-accuracy modes carry higher per-page rates than text-layer modes.
There is no monthly cap. Cost scales directly with the number of pages you process in each mode. You pay only for what you use, with no upfront commitment. API-level rate limits exist to keep the service fair, but sit above typical production needs.
unstract.com
Helpful?
Vendor refund policy
Usage charges are for pages already processed and are generally not refundable. If you believe you were billed in error, contact support@unstract.com within 30 days of the charge with your AWS account ID and the billing period, and we will review the request.
Legal
Vendor terms and conditions
Upon subscribing to this product, you must acknowledge and agree to the terms and conditions outlined in the vendor's End User License Agreement (EULA) .
Content disclaimer
Vendors are responsible for their product descriptions and other product content. AWS does not warrant that vendors' product descriptions or other product content are accurate, complete, reliable, current, or error-free.
Delivery details
64-bit (x86) Amazon Machine Image (AMI)
Amazon Machine Image (AMI)
An AMI is a virtual image that provides the information required to launch an instance. Amazon EC2 (Elastic Compute Cloud) instances are virtual servers on which you can run your applications and workloads, offering varying combinations of CPU, memory, storage, and networking resources. You can launch as many instances from as many different AMIs as you need.
Version release notes
LLMWhisperer v2.65.1: first public release of the self-hosted LLMWhisperer appliance on AWS Marketplace.
LLMWhisperer turns PDFs, scanned documents, images and Excel files into layout-preserving text for LLMs and RAG pipelines. The complete stack runs on a single EC2 instance in your AWS account, so your documents never leave the instance.
Included in this release:
- Five extraction modes, each billed per page: native_text, low_cost, high_quality, form (including XLS/XLSX) and table.
- REST API (v2) on port 3006, with an API key generated for each instance on first boot.
- Web dashboard on port 3007 showing license status and usage, behind a sign-in.
- Automatic licensing from your AWS Marketplace subscription. No manual activation.
- Hourly per-page usage reporting to AWS Marketplace, billed on your AWS invoice.
- Automatic sizing: on first boot the appliance sizes its processing to the instance's CPU and memory.
- Two launch options: the AMI from the EC2 console, or a CloudFormation template that creates the instance, IAM role and security group. Access to the API and dashboard, and optionally SSH, is restricted to the CIDR ranges you choose.
- A built-in status command to follow first boot and check the appliance's health.
Setup, networking requirements and troubleshooting are covered in the deployment guide: https://docs.unstract.com/llmwhisperer/llm_whisperer/aws_marketplace/aws_marketplace_deployment/
Additional details
Usage instructions
- Launch
- Launch from the EC2 console ("Launch from EC2 Console" on the subscription page). Do NOT use one-click launch: it does not attach an IAM instance role, and without one the appliance cannot obtain its license and does not start.
- Attach an IAM instance role that allows aws-marketplace:MeterUsage.
- Recommended instance type: m5.2xlarge or larger. Choose the size at launch: the appliance sizes its extraction engines for that instance type on first boot, so to use a different size, launch a new instance instead of changing the type of this one.
- Security group: allow TCP 3006 (API) and 3007 (dashboard) only from trusted addresses. Allow TCP 22 only if you use SSH. Session Manager needs no inbound port, but the instance role must also allow Session Manager if you want to use it.
- Network requirements (ongoing) This product requires an ongoing outbound HTTPS (TCP 443) connection. The following external services are required to function:
- Unstract license service: issues the license at first boot and renews it automatically. Without access, high_quality, form and table stop after roughly 100-150 minutes; native_text and low_cost work until license expiry.
- AWS Marketplace Metering Service: receives the hourly page counts. The instance also reads its role credentials from the EC2 instance metadata service. Nothing is downloaded at deployment; all software is included in the AMI.
-
First boot First boot can take 15 minutes or more while the license is obtained and the extraction engines load. To check progress, connect with SSH (user ec2-user) or Session Manager and run:
sudo python3 /usr/local/lib/llmw/llmw_provision.py --status
Look for "state": "provisioned" and "stack": "running", then confirm HTTP 200 from http://127.0.0.1:3006/health/ping on the instance. The first high_quality, form or table request after startup can take several minutes while OCR engines initialise. If the license is refused, the status output explains why. The appliance retries automatically.
-
Credentials Each instance generates its own API key and dashboard login, stored only in this file. Read them with:
sudo cat /opt/llmwhisperer/credentials
-
Call the API Send the unstract-key header from the credentials file on every request:
curl -X POST "http://<instance-address>:3006/api/v2/whisper?mode=native_text&output_mode=layout_preserving"
-H "unstract-key: <key>"
-H "Content-Type: application/pdf"
--data-binary @document.pdf
The response contains a whisper_hash. Check progress, then fetch the text:
curl "http://<instance-address>:3006/api/v2/whisper-status?whisper_hash=<hash>" -H "unstract-key: <key>" curl "http://<instance-address>:3006/api/v2/whisper-retrieve?whisper_hash=<hash>" -H "unstract-key: <key>"Supported modes: native_text, low_cost, high_quality, form, table.
The appliance serves the same v2 API as LLMWhisperer cloud. Use your instance address in place of the cloud URL:
- AWS Marketplace deployment guide: https://docs.unstract.com/llmwhisperer/llm_whisperer/aws_marketplace/aws_marketplace_deployment/
- API reference: https://docs.unstract.com/llmwhisperer/llm_whisperer/apis/index/
- Postman collection: https://docs.unstract.com/llmwhisperer/llm_whisperer/apis/llm_whisperer_postman_collection/ After importing, set baseUrl to http://<instance-address>:3006/api/v2 and apiKey to your key.
- Webhooks: https://docs.unstract.com/llmwhisperer/llm_whisperer/llm_whisperer_webhooks/
-
Dashboard Open http://<instance-address>:3007/ and sign in with the username and password from the credentials file. Older images lack auth on usage/license routes; 0.0.0.0/0 exposes them to all IPv4 addresses over plain HTTP.
-
Billing You are billed per page processed, depending on the mode you use. Excel files are billed at the form mode rate. EC2 charges are billed separately.
Resources
Vendor resources
Support
Vendor support
If you need assistance or face any issues during use, please contact us at support@unstract.com . Please include your AWS account ID, the instance ID, the output of 'sudo python3 /usr/local/lib/llmw/llmw_provision.py --status', and a description of the issue.
AWS infrastructure support
AWS Support is a one-on-one, fast-response support channel that is staffed 24x7x365 with experienced and technical support engineers. The service helps customers of all sizes and technical abilities to successfully utilize the products and features provided by Amazon Web Services.
Similar products
![LLMWhisperer [US] - Layout-Preserving Document Parser for LLM-Ready Text](https://d7umqicpi7263.cloudfront.net/img/product/796f5ebe-f095-4bef-bec9-ca2268933204.com/0c60012ebe21dad79514b9e6c14c077a)
LLMWhisperer converts complex documents into clean, layout-preserving text that large language models can read.
![LLMWhisperer [EU] - Layout-Preserving Document Parser for LLM-Ready Text](https://d7umqicpi7263.cloudfront.net/img/product/2a3db97a-6445-430d-84e2-e32cdf595959.com/0c60012ebe21dad79514b9e6c14c077a)
LLMWhisperer converts complex documents into clean, layout-preserving text that large language models can read.
REQUIRES PRIVATE OFFER
To purchase LiteLLM Enterprise Self-Hosted, please reach out to sales@berri.ai for a Private Offer.
LiteLLM is an OpenAI compatible Proxy Server (LLM Gateway) to call 2,000+ LLM APIs using the OpenAI format Bedrock, Huggingface, VertexAI, TogetherAI, Azure OpenAI, OpenAI, etc. Get started with Opensource LiteLLM here: https://github.com/BerriAI/litellm (40,000+ Github Stars)
This is a repackaged open source software product wherein additional charges apply for hardening, security configuration, and setup support.
vLLM is a high-throughput OpenAI-compatible inference server for open-source LLMs (Python + PyTorch CPU build), bundled with Open WebUI as a browser chat front end. This Lynxroute build is secured and ready out of the box: a random 32-byte API key generated at first boot, Bearer-token auth enforced on every /v1 request, vLLM and Open WebUI bound to loopback behind Nginx TLS, a default tiny model (facebook/opt-125m) preloaded so the API and chat work immediately, UFW firewall pre-configured, and a CIS Level 1 hardened Ubuntu 24.04 LTS base.
vLLM is Apache-2.0 licensed; Open WebUI uses a source-available license (see Long Description for details).
Customer reviews
No customer reviews yet
Be the first to review this product . We've partnered with PeerSpot to gather customer feedback. You can share your experience by writing or recording a review, or scheduling a call with a PeerSpot analyst.