Logistics Cash Flow & Invoice Reconciliation Pipeline using LangChain on Modal
Supply chain vendors face severe cash flow bottlenecks due to manual invoice and proof-of-delivery reconciliation delays. Build a distributed LangChain pipeline deployed on Modal to process receipt documents and reconcile ledger discrepancies, achieving 90% automated invoice matching accuracy under 500ms serverless execution.
What you are building
The core problem, expected build, and operating context for this challenge.
Build a serverless invoice and proof-of-delivery reconciliation pipeline using LangChain agents containerized and executed on Modal.
How work is evaluated
Evaluates matching accuracy between supply chain invoice JSON inputs and purchase ledger baseline.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
strict_signature_check
Ensures missing POD signatures automatically flag DISCREPANCY status.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
matching_accuracy
Accuracy of invoice reconciliation status. • target: 0.9 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Construct modular document extraction chains using LangChain
Deploy containerized LLM tasks on Modal for fast serverless execution
Automate reconciliation between purchase orders, delivery notes, and tax invoices
Calculate variance metrics and route mismatched records for manual review
Reference links and supporting material
250 logistics purchase order and vendor invoice mock pairs.
Challenge setup
Evaluates matching accuracy between supply chain invoice JSON inputs and purchase ledger baseline.
Challenge input
JSON object with po_number, invoice_total, line_items array, and pod_signature_present bool.
LangChain
Provides LCEL abstractions for prompt and tool chaining.
Modal
Enables fast, scalable Python microservice execution.
Evaluated output
JSON object with status ('MATCHED', 'DISCREPANCY', 'REJECTED'), total_diff, and resolution_notes.
- Ensures missing POD signatures automatically flag DISCREPANCY status.
- Accuracy of invoice reconciliation status. • target: 0.9 • range: 0-1
- Benchmark: Document Extraction & Accounting Verification Benchmark
- Matching Accuracy target: 0.9
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- Modal · Optional
- LangChain · Required
- Langchain · Optional
- Modal · Optional
- LangChain · Required
- Langchain · Optional
Evaluation contract
- strict_signature_check · Weight 1
- matching_accuracy · Weight 1
Metrics
- matching_accuracy · Target 0.9
Benchmarks
- Document Extraction & Accounting Verification Benchmark
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON object with po_number, invoice_total, line_items array, and pod_signature_present bool.
Output: JSON object with status ('MATCHED', 'DISCREPANCY', 'REJECTED'), total_diff, and resolution_notes.
Execution harness: Python
Evaluation tools
- LangChain · Chain Architecture
- Modal · Serverless Runtime
Recipe freeze marker
No recipe freeze time is recorded.