my-project-public

repository

loading code, commits, and activity

repositories

loading repo index

my-project-public — gitlawb

#1	---
#2	title: Cohere
#3	description: "Reranking with Cohere"
#4	---
#5
#6	Cohere provides enterprise-grade reranking models with excellent multilingual support and production-ready performance.
#7
#8	## Models
#9
#10	Cohere offers several reranking models:
#11
#12	- `rerank-english-v3.0`: Latest English reranker with best performance
#13	- `rerank-multilingual-v3.0`: Multilingual support for global applications
#14	- `rerank-english-v2.0`: Previous generation English reranker
#15
#16	## Installation
#17
#18	```bash
#19	pip install cohere
#20	```
#21
#22	## Configuration
#23
#24	```python Python
#25	from mem0 import Memory
#26
#27	config = {
#28	"vector_store": {
#29	"provider": "chroma",
#30	"config": {
#31	"collection_name": "my_memories",
#32	"path": "./chroma_db"
#33	}
#34	},
#35	"llm": {
#36	"provider": "openai",
#37	"config": {
#38	"model": "gpt-4.1-nano-2025-04-14"
#39	}
#40	},
#41	"reranker": {
#42	"provider": "cohere",
#43	"config": {
#44	"model": "rerank-english-v3.0",
#45	"api_key": "your-cohere-api-key", # or set COHERE_API_KEY
#46	"top_k": 5,
#47	"return_documents": False,
#48	"max_chunks_per_doc": None
#49	}
#50	}
#51	}
#52
#53	memory = Memory.from_config(config)
#54	```
#55
#56	## Environment Variables
#57
#58	Set your API key as an environment variable:
#59
#60	```bash
#61	export COHERE_API_KEY="your-api-key"
#62	```
#63
#64	## Usage Example
#65
#66	```python Python
#67	import os
#68	from mem0 import Memory
#69
#70	# Set API key
#71	os.environ["COHERE_API_KEY"] = "your-api-key"
#72
#73	# Initialize memory with Cohere reranker
#74	config = {
#75	"vector_store": {"provider": "chroma"},
#76	"llm": {"provider": "openai", "config": {"model": "gpt-4o-mini"}},
#77	"rerank": {
#78	"provider": "cohere",
#79	"config": {
#80	"model": "rerank-english-v3.0",
#81	"top_k": 3
#82	}
#83	}
#84	}
#85
#86	memory = Memory.from_config(config)
#87
#88	# Add memories
#89	messages = [
#90	{"role": "user", "content": "I work as a data scientist at Microsoft"},
#91	{"role": "user", "content": "I specialize in machine learning and NLP"},
#92	{"role": "user", "content": "I enjoy playing tennis on weekends"}
#93	]
#94
#95	memory.add(messages, user_id="bob")
#96
#97	# Search with reranking
#98	results = memory.search("What is the user's profession?", user_id="bob")
#99
#100	for result in results['results']:
#101	print(f"Memory: {result['memory']}")
#102	print(f"Vector Score: {result['score']:.3f}")
#103	print(f"Rerank Score: {result['rerank_score']:.3f}")
#104	print()
#105	```
#106
#107	## Multilingual Support
#108
#109	For multilingual applications, use the multilingual model:
#110
#111	```python Python
#112	config = {
#113	"rerank": {
#114	"provider": "cohere",
#115	"config": {
#116	"model": "rerank-multilingual-v3.0",
#117	"top_k": 5
#118	}
#119	}
#120	}
#121	```
#122
#123	## Configuration Parameters
#124
#125	\| Parameter \| Description \| Type \| Default \|
#126	\| -------------------- \| -------------------------------- \| ------ \| ----------------------- \|
#127	\| `model` \| Cohere rerank model to use \| `str` \| `"rerank-english-v3.0"` \|
#128	\| `api_key` \| Cohere API key \| `str` \| `None` \|
#129	\| `top_k` \| Maximum documents to return \| `int` \| `None` \|
#130	\| `return_documents` \| Whether to return document texts \| `bool` \| `False` \|
#131	\| `max_chunks_per_doc` \| Maximum chunks per document \| `int` \| `None` \|
#132
#133	## Features
#134
#135	- High Quality: Enterprise-grade relevance scoring
#136	- Multilingual: Support for 100+ languages
#137	- Scalable: Production-ready with high throughput
#138	- Reliable: SLA-backed service with 99.9% uptime
#139
#140	## Best Practices
#141
#142	1. Model Selection: Use `rerank-english-v3.0` for English, `rerank-multilingual-v3.0` for other languages
#143	2. Batch Processing: Process multiple queries efficiently
#144	3. Error Handling: Implement retry logic for production systems
#145	4. Monitoring: Track reranking performance and costs
#146

z6Mkq5mY3JWtxoxUobWcfNHm7AkRubgSWEZTkBVqZXJviFZ5/my-project-public