| Name | ai-batch JSON |
| Version |
0.2.0
JSON |
| download |
| home_page | None |
| Summary | Batch processing for Anthropic's Claude API with structured output |
| upload_time | 2025-07-11 04:22:25 |
| maintainer | None |
| docs_url | None |
| author | Agam More |
| requires_python | >=3.12 |
| license | MIT |
| keywords |
|
| VCS |
|
| bugtrack_url |
|
| requirements |
No requirements were recorded.
|
| Travis-CI |
No Travis.
|
| coveralls test coverage |
No coveralls.
|
# AI Batch
Python SDK for **batch processing** with structured output and citation mapping.
- **50% cost savings** via Anthropic's batch API pricing
- **Automatic cost tracking** with token usage and pricing
- **Structured output** with Pydantic models
- **Field-level citations** map results to source documents
- **Type safety** with full validation
Currently supports Anthropic Claude. OpenAI support coming soon.
## API Reference
- [`batch()`](#batch) - Process message conversations or PDF files
- [`BatchJob`](#batchjob) - Job status and results
- [`BatchManager`](#batchmanager) - Manage large-scale batch processing with parallel execution
## Quick Start
```python
from ai_batch import batch
from pydantic import BaseModel
class Invoice(BaseModel):
company_name: str
total_amount: str
date: str
# Process PDFs with structured output + citations
job = batch(
files=["invoice1.pdf", "invoice2.pdf", "invoice3.pdf"],
prompt="Extract the company name, total amount, and date.",
model="claude-3-5-sonnet-20241022",
response_model=Invoice,
enable_citations=True
)
# Wait for completion
while not job.is_complete():
time.sleep(30)
results = job.results()
# Results now contain both data and citations together:
# [{"result": Invoice(...), "citations": {"company_name": [Citation(...)], ...}}, ...]
```
## Installation
```bash
pip install ai-batch
```
## Usage
Create a `.env` file in your project root:
```bash
ANTHROPIC_API_KEY=your-api-key
```
## API Functions
### batch()
Process multiple message conversations with optional structured output.
```python
from ai_batch import batch
from pydantic import BaseModel
class SpamResult(BaseModel):
is_spam: bool
confidence: float
reason: str
# Process messages
job = batch(
messages=[
[{"role": "user", "content": "Is this spam? You've won $1000!"}],
[{"role": "user", "content": "Meeting at 3pm tomorrow"}],
[{"role": "user", "content": "URGENT: Click here now!"}]
],
model="claude-3-haiku-20240307",
response_model=SpamResult
)
# Get results
results = job.results()
```
**Response:**
```python
[
SpamResult(is_spam=True, confidence=0.95, reason="Contains monetary prize claim"),
SpamResult(is_spam=False, confidence=0.98, reason="Normal meeting reminder"),
SpamResult(is_spam=True, confidence=0.92, reason="Urgent call-to-action pattern")
]
```
### batch() with files
Process PDF files with optional structured output and citations.
```python
from ai_batch import batch
from pydantic import BaseModel
class Invoice(BaseModel):
company_name: str
total_amount: str
date: str
# Process PDFs with citations
job = batch(
files=["invoice1.pdf", "invoice2.pdf"],
prompt="Extract the company name, total amount, and date.",
model="claude-3-5-sonnet-20241022",
response_model=Invoice,
enable_citations=True
)
results = job.results()
# Results now contain both data and citations together
```
**Response:**
```python
# Results now contain both data and citations together
[
{
"result": Invoice(company_name="TechCorp Solutions", total_amount="$12,500.00", date="March 15, 2024"),
"citations": {
"company_name": [Citation(cited_text="TechCorp Solutions", start_page=1)],
"total_amount": [Citation(cited_text="TOTAL: $12,500.00", start_page=2)],
"date": [Citation(cited_text="Date: March 15, 2024", start_page=1)]
}
},
{
"result": Invoice(company_name="DataFlow Systems", total_amount="$8,750.00", date="March 18, 2024"),
"citations": {
"company_name": [Citation(cited_text="DataFlow Systems", start_page=1)],
"total_amount": [Citation(cited_text="Total Due: $8,750.00", start_page=3)],
"date": [Citation(cited_text="Invoice Date: March 18, 2024", start_page=1)]
}
}
]
```
### BatchJob
The job object returned by `batch()`.
```python
# Check completion status
if job.is_complete():
results = job.results()
# Get processing statistics with cost tracking
stats = job.stats(print_stats=True)
# Output:
# 📊 Batch Statistics
# ID: msgbatch_01BPtdnmEwxtaDcdJ2eUsq4T
# Status: ended
# Complete: ✅
# Elapsed: 41.8s
# Mode: Text + Citations
# Results: 0
# Citations: 0
# Input tokens: 2,117
# Output tokens: 81
# Total cost: $0.0038
# (50% batch discount applied)
# Citations are now included in results (if enabled)
# Access via: results[0]["citations"]
# Save raw API responses
job = batch(..., raw_results_dir="./raw_responses")
```
### BatchManager
Manage large-scale batch processing with automatic job splitting, parallel execution, state persistence, and cost management.
```python
from ai_batch import BatchManager
from pydantic import BaseModel
class Invoice(BaseModel):
company_name: str
total_amount: float
invoice_number: str
# Initialize BatchManager for large-scale processing
manager = BatchManager(
files=["invoice1.pdf", "invoice2.pdf", ...], # 100+ files
prompt="Extract invoice data",
model="claude-3-5-sonnet-20241022",
response_model=Invoice,
enable_citations=True,
items_per_job=10, # Process 10 files per job
max_parallel_jobs=5, # 5 jobs in parallel
max_cost=50.0, # Stop if cost exceeds $50
state_path="batch_state.json", # Auto-resume capability
save_results_dir="results/" # Save results to disk
)
# Run processing (auto-resumes if interrupted)
summary = manager.run(print_progress=True)
# Retry failed items
if summary['failed_items'] > 0:
retry_summary = manager.retry_failed()
# Get statistics
stats = manager.stats
print(f"Completed: {stats['completed_items']}/{stats['total_items']}")
print(f"Total cost: ${stats['total_cost']:.2f}")
# Load results from disk
results = manager.get_results_from_disk()
```
**Key Features:**
- **Automatic job splitting**: Breaks large batches into smaller chunks
- **Parallel processing**: Multiple jobs run concurrently with ThreadPoolExecutor
- **State persistence**: Resume from interruptions with JSON state files
- **Cost management**: Stop processing when budget limit is reached
- **Progress monitoring**: Real-time progress updates with statistics
- **Retry mechanism**: Easily retry failed items
- **Result saving**: Organized directory structure for results
## Citations
Citations work in two modes depending on whether you use structured output:
### 1. Text + Citations (Flat List)
When `enable_citations=True` without a response model, citations are returned as a flat list:
```python
job = batch(
files=["document.pdf"],
prompt="Summarize the key findings",
enable_citations=True
)
results = job.results() # List of {"result": str, "citations": List[Citation]}
# Example result structure:
[
{
"result": "Summary text...",
"citations": [
Citation(cited_text="AI reduces errors by 30%", start_page=2),
Citation(cited_text="Implementation cost: $50,000", start_page=5)
]
}
]
```
### 2. Structured + Field Citations (Mapping)
When using both `response_model` and `enable_citations=True`, citations are mapped to specific fields:
```python
job = batch(
files=["document.pdf"],
prompt="Extract the data",
response_model=MyModel,
enable_citations=True
)
results = job.results() # List of {"result": Model, "citations": Dict[str, List[Citation]]}
# Example result structure:
[
{
"result": MyModel(title="Annual Report 2024", revenue="$1.2M"),
"citations": {
"title": [Citation(cited_text="Annual Report 2024", start_page=1)],
"revenue": [Citation(cited_text="Revenue: $1.2M", start_page=3)],
"growth": [Citation(cited_text="YoY Growth: 25%", start_page=3)]
}
}
]
```
The field mapping allows you to trace exactly which part of the source document was used to populate each field in your structured output.
### Robust Citation Parsing
AI Batch uses proper JSON parsing for citation field mapping, ensuring reliability with complex JSON structures:
**Handles Complex Scenarios:**
- ✅ Escaped quotes in JSON values: `"name": "John \"The Great\" Doe"`
- ✅ URLs with colons: `"website": "http://example.com:8080"`
- ✅ Nested objects and arrays: `"metadata": {"nested": {"deep": "value"}}`
- ✅ Multi-line strings and special characters
- ✅ Fields with numbers/underscores: `user_name`, `age_2`
**Previous Limitations (Fixed):**
The old regex-based approach would fail on complex JSON patterns. The new JSON parser reliably handles any valid JSON structure that Claude produces, making citation mapping robust for production use.
## Cost Tracking
AI Batch automatically tracks token usage and costs for all batch operations:
```python
from ai_batch import batch
job = batch(
messages=[...],
model="claude-3-5-sonnet-20241022"
)
# Get cost information
stats = job.stats()
print(f"Total cost: ${stats['total_cost']:.4f}")
print(f"Input tokens: {stats['total_input_tokens']:,}")
print(f"Output tokens: {stats['total_output_tokens']:,}")
# Or print formatted statistics
job.stats(print_stats=True)
```
## Example Scripts
- [`examples/spam_detection.py`](examples/spam_detection.py) - Email classification
- [`examples/pdf_extraction.py`](examples/pdf_extraction.py) - PDF data extraction
- [`examples/citation_example.py`](examples/citation_example.py) - Basic citation usage
- [`examples/citation_with_pydantic.py`](examples/citation_with_pydantic.py) - Structured output with citations
- [`examples/batch_manager_example.py`](examples/batch_manager_example.py) - Large-scale batch processing with BatchManager
## Limitations
- Citationm mapping only work with flat Pydantic models (no nested models)
- No support for OpenAI.
- PDFs require Opus/Sonnet models for best results
- Batch jobs can take up to 24 hours to process
- Use `job.is_complete()` to check status before getting results
- Citations may not be available in all batch API responses
## Comparison with Alternatives
| Feature | ai-batch | LangChain | Instructor | PydanticAI |
|---------|----------|-----------|------------|------------|
| **Batch Requests** | ✅ Native (50% cost savings) | ❌ No native batch API | ✅ Via OpenAI Batch API ([#1092](https://github.com/instructor-ai/instructor/issues/1092)) | ⚠️ Planned ([#1771](https://github.com/pydantic/pydantic-ai/issues/1771)) |
| **Structured Output** | ✅ Full support | ✅ Via parsers | ✅ Core feature | ✅ Native |
| **PDF File Input** | ✅ Native support | ✅ Via document loaders | ✅ Via multimodal models | ✅ Via file handling |
| **Citation Mapping** | ✅ Field-level citations | ❌ Manual implementation | ❌ Manual implementation | ❌ Manual implementation |
| **Cost Tracking** | ✅ Automatic with tokencost | ❌ Manual implementation | ❌ Manual implementation | ❌ Manual implementation |
| **Cost Limits** | ✅ max_cost parameter | ❌ Manual implementation | ❌ Manual implementation | ❌ Manual implementation |
| **Batch Providers** | 2/2 (Anthropic, OpenAI planned) | 0/2 | 1/2 (OpenAI only) | 0/2 |
| **Focus** | Streamlined batch requests | General LLM orchestration | Structured outputs CLI | Agent framework |
## License
MIT
## Todos
- [x] ~~Add pricing metadata and max_spend controls~~ (Cost tracking implemented)
- [x] ~~Auto batch manager (parallel batches, retry, spend control)~~ (BatchManager implemented)
- [ ] Test mode to run on 1% sample before full batch
- [ ] Quick batch - split into smaller chunks for faster results
- [ ] Support text/other file types (not just PDFs)
- [ ] Support for OpenAI
Raw data
{
"_id": null,
"home_page": null,
"name": "ai-batch",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.12",
"maintainer_email": null,
"keywords": null,
"author": "Agam More",
"author_email": null,
"download_url": "https://files.pythonhosted.org/packages/41/c3/f24582405bf70090136bde085ef2d0aa1afa2771a02fefca46fd5c0c9817/ai_batch-0.2.0.tar.gz",
"platform": null,
"description": "# AI Batch\n\nPython SDK for **batch processing** with structured output and citation mapping.\n\n- **50% cost savings** via Anthropic's batch API pricing\n- **Automatic cost tracking** with token usage and pricing\n- **Structured output** with Pydantic models \n- **Field-level citations** map results to source documents\n- **Type safety** with full validation\n\nCurrently supports Anthropic Claude. OpenAI support coming soon.\n\n## API Reference\n\n- [`batch()`](#batch) - Process message conversations or PDF files\n- [`BatchJob`](#batchjob) - Job status and results\n- [`BatchManager`](#batchmanager) - Manage large-scale batch processing with parallel execution\n\n## Quick Start\n\n```python\nfrom ai_batch import batch\nfrom pydantic import BaseModel\n\nclass Invoice(BaseModel):\n company_name: str\n total_amount: str\n date: str\n\n# Process PDFs with structured output + citations\njob = batch(\n files=[\"invoice1.pdf\", \"invoice2.pdf\", \"invoice3.pdf\"],\n prompt=\"Extract the company name, total amount, and date.\",\n model=\"claude-3-5-sonnet-20241022\",\n response_model=Invoice,\n enable_citations=True\n)\n\n# Wait for completion\nwhile not job.is_complete():\n time.sleep(30)\n \nresults = job.results()\n# Results now contain both data and citations together:\n# [{\"result\": Invoice(...), \"citations\": {\"company_name\": [Citation(...)], ...}}, ...]\n```\n\n## Installation\n\n```bash\npip install ai-batch\n```\n\n## Usage\n\nCreate a `.env` file in your project root:\n\n```bash\nANTHROPIC_API_KEY=your-api-key\n```\n\n## API Functions\n\n### batch()\n\nProcess multiple message conversations with optional structured output.\n\n```python\nfrom ai_batch import batch\nfrom pydantic import BaseModel\n\nclass SpamResult(BaseModel):\n is_spam: bool\n confidence: float\n reason: str\n\n# Process messages\njob = batch(\n messages=[\n [{\"role\": \"user\", \"content\": \"Is this spam? You've won $1000!\"}],\n [{\"role\": \"user\", \"content\": \"Meeting at 3pm tomorrow\"}],\n [{\"role\": \"user\", \"content\": \"URGENT: Click here now!\"}]\n ],\n model=\"claude-3-haiku-20240307\",\n response_model=SpamResult\n)\n\n# Get results\nresults = job.results()\n```\n\n**Response:**\n```python\n[\n SpamResult(is_spam=True, confidence=0.95, reason=\"Contains monetary prize claim\"),\n SpamResult(is_spam=False, confidence=0.98, reason=\"Normal meeting reminder\"),\n SpamResult(is_spam=True, confidence=0.92, reason=\"Urgent call-to-action pattern\")\n]\n```\n\n### batch() with files\n\nProcess PDF files with optional structured output and citations.\n\n```python\nfrom ai_batch import batch\nfrom pydantic import BaseModel\n\nclass Invoice(BaseModel):\n company_name: str\n total_amount: str\n date: str\n\n# Process PDFs with citations\njob = batch(\n files=[\"invoice1.pdf\", \"invoice2.pdf\"],\n prompt=\"Extract the company name, total amount, and date.\",\n model=\"claude-3-5-sonnet-20241022\",\n response_model=Invoice,\n enable_citations=True\n)\n\nresults = job.results()\n# Results now contain both data and citations together\n```\n\n**Response:**\n```python\n# Results now contain both data and citations together\n[\n {\n \"result\": Invoice(company_name=\"TechCorp Solutions\", total_amount=\"$12,500.00\", date=\"March 15, 2024\"),\n \"citations\": {\n \"company_name\": [Citation(cited_text=\"TechCorp Solutions\", start_page=1)],\n \"total_amount\": [Citation(cited_text=\"TOTAL: $12,500.00\", start_page=2)],\n \"date\": [Citation(cited_text=\"Date: March 15, 2024\", start_page=1)]\n }\n },\n {\n \"result\": Invoice(company_name=\"DataFlow Systems\", total_amount=\"$8,750.00\", date=\"March 18, 2024\"),\n \"citations\": {\n \"company_name\": [Citation(cited_text=\"DataFlow Systems\", start_page=1)],\n \"total_amount\": [Citation(cited_text=\"Total Due: $8,750.00\", start_page=3)],\n \"date\": [Citation(cited_text=\"Invoice Date: March 18, 2024\", start_page=1)]\n }\n }\n]\n```\n\n### BatchJob\n\nThe job object returned by `batch()`.\n\n```python\n# Check completion status\nif job.is_complete():\n results = job.results()\n\n# Get processing statistics with cost tracking\nstats = job.stats(print_stats=True)\n# Output:\n# \ud83d\udcca Batch Statistics\n# ID: msgbatch_01BPtdnmEwxtaDcdJ2eUsq4T\n# Status: ended\n# Complete: \u2705\n# Elapsed: 41.8s\n# Mode: Text + Citations\n# Results: 0\n# Citations: 0\n# Input tokens: 2,117\n# Output tokens: 81\n# Total cost: $0.0038\n# (50% batch discount applied)\n\n# Citations are now included in results (if enabled)\n# Access via: results[0][\"citations\"]\n\n# Save raw API responses\njob = batch(..., raw_results_dir=\"./raw_responses\")\n```\n\n### BatchManager\n\nManage large-scale batch processing with automatic job splitting, parallel execution, state persistence, and cost management.\n\n```python\nfrom ai_batch import BatchManager\nfrom pydantic import BaseModel\n\nclass Invoice(BaseModel):\n company_name: str\n total_amount: float\n invoice_number: str\n\n# Initialize BatchManager for large-scale processing\nmanager = BatchManager(\n files=[\"invoice1.pdf\", \"invoice2.pdf\", ...], # 100+ files\n prompt=\"Extract invoice data\",\n model=\"claude-3-5-sonnet-20241022\",\n response_model=Invoice,\n enable_citations=True,\n items_per_job=10, # Process 10 files per job\n max_parallel_jobs=5, # 5 jobs in parallel\n max_cost=50.0, # Stop if cost exceeds $50\n state_path=\"batch_state.json\", # Auto-resume capability\n save_results_dir=\"results/\" # Save results to disk\n)\n\n# Run processing (auto-resumes if interrupted)\nsummary = manager.run(print_progress=True)\n\n# Retry failed items\nif summary['failed_items'] > 0:\n retry_summary = manager.retry_failed()\n\n# Get statistics\nstats = manager.stats\nprint(f\"Completed: {stats['completed_items']}/{stats['total_items']}\")\nprint(f\"Total cost: ${stats['total_cost']:.2f}\")\n\n# Load results from disk\nresults = manager.get_results_from_disk()\n```\n\n**Key Features:**\n- **Automatic job splitting**: Breaks large batches into smaller chunks\n- **Parallel processing**: Multiple jobs run concurrently with ThreadPoolExecutor\n- **State persistence**: Resume from interruptions with JSON state files\n- **Cost management**: Stop processing when budget limit is reached\n- **Progress monitoring**: Real-time progress updates with statistics\n- **Retry mechanism**: Easily retry failed items\n- **Result saving**: Organized directory structure for results\n\n## Citations\n\nCitations work in two modes depending on whether you use structured output:\n\n### 1. Text + Citations (Flat List)\n\nWhen `enable_citations=True` without a response model, citations are returned as a flat list:\n\n```python\njob = batch(\n files=[\"document.pdf\"],\n prompt=\"Summarize the key findings\",\n enable_citations=True\n)\n\nresults = job.results() # List of {\"result\": str, \"citations\": List[Citation]}\n\n# Example result structure:\n[\n {\n \"result\": \"Summary text...\",\n \"citations\": [\n Citation(cited_text=\"AI reduces errors by 30%\", start_page=2),\n Citation(cited_text=\"Implementation cost: $50,000\", start_page=5)\n ]\n }\n]\n```\n\n### 2. Structured + Field Citations (Mapping)\n\nWhen using both `response_model` and `enable_citations=True`, citations are mapped to specific fields:\n\n```python\njob = batch(\n files=[\"document.pdf\"],\n prompt=\"Extract the data\",\n response_model=MyModel,\n enable_citations=True\n)\n\nresults = job.results() # List of {\"result\": Model, \"citations\": Dict[str, List[Citation]]}\n\n# Example result structure:\n[\n {\n \"result\": MyModel(title=\"Annual Report 2024\", revenue=\"$1.2M\"),\n \"citations\": {\n \"title\": [Citation(cited_text=\"Annual Report 2024\", start_page=1)],\n \"revenue\": [Citation(cited_text=\"Revenue: $1.2M\", start_page=3)],\n \"growth\": [Citation(cited_text=\"YoY Growth: 25%\", start_page=3)]\n }\n }\n]\n```\n\nThe field mapping allows you to trace exactly which part of the source document was used to populate each field in your structured output.\n\n### Robust Citation Parsing\n\nAI Batch uses proper JSON parsing for citation field mapping, ensuring reliability with complex JSON structures:\n\n**Handles Complex Scenarios:**\n- \u2705 Escaped quotes in JSON values: `\"name\": \"John \\\"The Great\\\" Doe\"`\n- \u2705 URLs with colons: `\"website\": \"http://example.com:8080\"`\n- \u2705 Nested objects and arrays: `\"metadata\": {\"nested\": {\"deep\": \"value\"}}`\n- \u2705 Multi-line strings and special characters\n- \u2705 Fields with numbers/underscores: `user_name`, `age_2`\n\n**Previous Limitations (Fixed):**\nThe old regex-based approach would fail on complex JSON patterns. The new JSON parser reliably handles any valid JSON structure that Claude produces, making citation mapping robust for production use.\n\n## Cost Tracking\n\nAI Batch automatically tracks token usage and costs for all batch operations:\n\n```python\nfrom ai_batch import batch\n\njob = batch(\n messages=[...],\n model=\"claude-3-5-sonnet-20241022\"\n)\n\n# Get cost information\nstats = job.stats()\nprint(f\"Total cost: ${stats['total_cost']:.4f}\")\nprint(f\"Input tokens: {stats['total_input_tokens']:,}\")\nprint(f\"Output tokens: {stats['total_output_tokens']:,}\")\n\n# Or print formatted statistics\njob.stats(print_stats=True)\n```\n\n## Example Scripts\n\n- [`examples/spam_detection.py`](examples/spam_detection.py) - Email classification\n- [`examples/pdf_extraction.py`](examples/pdf_extraction.py) - PDF data extraction\n- [`examples/citation_example.py`](examples/citation_example.py) - Basic citation usage\n- [`examples/citation_with_pydantic.py`](examples/citation_with_pydantic.py) - Structured output with citations\n- [`examples/batch_manager_example.py`](examples/batch_manager_example.py) - Large-scale batch processing with BatchManager\n\n## Limitations\n\n- Citationm mapping only work with flat Pydantic models (no nested models)\n- No support for OpenAI.\n- PDFs require Opus/Sonnet models for best results\n- Batch jobs can take up to 24 hours to process\n- Use `job.is_complete()` to check status before getting results\n- Citations may not be available in all batch API responses\n\n## Comparison with Alternatives\n\n| Feature | ai-batch | LangChain | Instructor | PydanticAI |\n|---------|----------|-----------|------------|------------|\n| **Batch Requests** | \u2705 Native (50% cost savings) | \u274c No native batch API | \u2705 Via OpenAI Batch API ([#1092](https://github.com/instructor-ai/instructor/issues/1092)) | \u26a0\ufe0f Planned ([#1771](https://github.com/pydantic/pydantic-ai/issues/1771)) |\n| **Structured Output** | \u2705 Full support | \u2705 Via parsers | \u2705 Core feature | \u2705 Native |\n| **PDF File Input** | \u2705 Native support | \u2705 Via document loaders | \u2705 Via multimodal models | \u2705 Via file handling |\n| **Citation Mapping** | \u2705 Field-level citations | \u274c Manual implementation | \u274c Manual implementation | \u274c Manual implementation |\n| **Cost Tracking** | \u2705 Automatic with tokencost | \u274c Manual implementation | \u274c Manual implementation | \u274c Manual implementation |\n| **Cost Limits** | \u2705 max_cost parameter | \u274c Manual implementation | \u274c Manual implementation | \u274c Manual implementation |\n| **Batch Providers** | 2/2 (Anthropic, OpenAI planned) | 0/2 | 1/2 (OpenAI only) | 0/2 |\n| **Focus** | Streamlined batch requests | General LLM orchestration | Structured outputs CLI | Agent framework |\n\n## License\n\nMIT\n\n## Todos\n\n- [x] ~~Add pricing metadata and max_spend controls~~ (Cost tracking implemented)\n- [x] ~~Auto batch manager (parallel batches, retry, spend control)~~ (BatchManager implemented)\n- [ ] Test mode to run on 1% sample before full batch\n- [ ] Quick batch - split into smaller chunks for faster results\n- [ ] Support text/other file types (not just PDFs)\n- [ ] Support for OpenAI",
"bugtrack_url": null,
"license": "MIT",
"summary": "Batch processing for Anthropic's Claude API with structured output",
"version": "0.2.0",
"project_urls": null,
"split_keywords": [],
"urls": [
{
"comment_text": null,
"digests": {
"blake2b_256": "8bc570ae5e07707ca68dbd77d0f28a1fc1759ab1da905139e11ae33bfdfae47c",
"md5": "4e403d9ffed275b446f0cfcc2c1ede1e",
"sha256": "1a144d25a601a293aa3dacd23cb44b50cbf5370035221cc1846a5a7dd069a6bb"
},
"downloads": -1,
"filename": "ai_batch-0.2.0-py3-none-any.whl",
"has_sig": false,
"md5_digest": "4e403d9ffed275b446f0cfcc2c1ede1e",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.12",
"size": 29789,
"upload_time": "2025-07-11T04:22:24",
"upload_time_iso_8601": "2025-07-11T04:22:24.462907Z",
"url": "https://files.pythonhosted.org/packages/8b/c5/70ae5e07707ca68dbd77d0f28a1fc1759ab1da905139e11ae33bfdfae47c/ai_batch-0.2.0-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": null,
"digests": {
"blake2b_256": "41c3f24582405bf70090136bde085ef2d0aa1afa2771a02fefca46fd5c0c9817",
"md5": "326fe31eadc33516c6339b017ed22f08",
"sha256": "8f43b285c6da70bc9bd38fa12926bf504d02cb84e2987830e8f031848a505932"
},
"downloads": -1,
"filename": "ai_batch-0.2.0.tar.gz",
"has_sig": false,
"md5_digest": "326fe31eadc33516c6339b017ed22f08",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.12",
"size": 125134,
"upload_time": "2025-07-11T04:22:25",
"upload_time_iso_8601": "2025-07-11T04:22:25.665269Z",
"url": "https://files.pythonhosted.org/packages/41/c3/f24582405bf70090136bde085ef2d0aa1afa2771a02fefca46fd5c0c9817/ai_batch-0.2.0.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-07-11 04:22:25",
"github": false,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"lcname": "ai-batch"
}
