Add llama-index-postprocessor-heroku-rerank integration

[dsouza-anush]

Summary

Adds llama-index-postprocessor-heroku-rerank package for Heroku Inference API reranking support.

Description

This integration enables LlamaIndex users to rerank retrieved documents using Heroku's managed inference service, which provides access to Cohere reranking models.

Features

• HerokuRerank class extending LlamaIndex BaseNodePostprocessor
• Support for cohere-rerank-3-5 model
• Configurable top_n for limiting reranked results
• Full sync and async support
• Seamless integration with LlamaIndex query engines

Usage

from llama_index.postprocessor.heroku_rerank import HerokuRerank

reranker = HerokuRerank(
    api_key="your-heroku-inference-key",
    model="cohere-rerank-3-5",
    top_n=5,
)

# Use with query engine
query_engine = index.as_query_engine(
    node_postprocessors=[reranker],
    similarity_top_k=20,  # Retrieve 20, rerank to top 5
)

Testing

• 10 unit tests with mocked HTTP responses
• Tested with real Heroku Inference API

Related

• Heroku Inference API: https://devcenter.heroku.com/articles/heroku-inference-api
• Existing llama-index-llms-heroku package
• PR Add llama-index-embeddings-heroku integration #20604 for embeddings integration

🤖 Generated with Claude Code

[AstraBert]

Looks like this pyproject is lacking many things and using poetry instead of hatchling + uv.

You should probably modify it taking this pyproject as reference:

[build-system]
requires = ["hatchling"]
build-backend = "hatchling.build"

[dependency-groups]
dev = [
    "black[jupyter]<=23.9.1,>=23.7.0",
    "codespell[toml]>=v2.2.6",
    "diff-cover>=9.2.0",
    "ipython==8.10.0",
    "jupyter>=1.0.0,<2",
    "mypy==0.991",
    "pre-commit==3.2.0",
    "pylint==2.15.10",
    "pytest==7.2.1",
    "pytest-asyncio>=0.23.8",
    "pytest-cov>=6.1.1",
    "pytest-mock==3.11.1",
    "ruff==0.11.11",
    "types-Deprecated>=0.1.0",
    "types-PyYAML>=6.0.12.12,<7",
    "types-protobuf>=4.24.0.4,<5",
    "types-redis==4.5.5.0",
    "types-requests==2.28.11.8",
    "types-setuptools==67.1.0.0",
]

[project]
name = "llama-index-llms-upstage"
version = "0.7.0"
description = "llama-index llms upstage integration"
authors = [{name = "Your Name", email = "you@example.com"}]
requires-python = ">=3.9,<4.0"
readme = "README.md"
license = "AGPL-3.0"
dependencies = [
    "tokenizers>=0.19.1,<0.20",
    "llama-index-core>=0.14.3,<0.15",
    "llama-index-llms-openai>=0.6.0,<0.7",
    "llama-index-readers-upstage>=0.4.0,<0.5",
]

[tool.codespell]
check-filenames = true
check-hidden = true
# Feel free to un-skip examples, and experimental, you will just need to
# work through many typos (--write-changes and --interactive will help)
skip = "*.csv,*.html,*.json,*.jsonl,*.pdf,*.txt,*.ipynb"

[tool.hatch.build.targets.sdist]
include = ["llama_index/"]

[tool.hatch.build.targets.wheel]
include = ["llama_index/"]

[tool.llamahub]
contains_example = false
import_path = "llama_index.llms.upstage"

[tool.llamahub.class_authors]
Upstage = "llama-index"

[tool.mypy]
disallow_untyped_defs = true
# Remove venv skip when integrated with pre-commit
exclude = ["_static", "build", "examples", "notebooks", "venv"]
ignore_missing_imports = true
python_version = "3.8"