git.iko.soy/crush
on main help sign in
docs(readme): simplify and cleanup readme
5004105Christian Rocha
README.md+13−68
compare with main
unifiedsplit
@@ -65,11 +65,9 @@ Crush can use LSPs for additional context to help inform its decisions, just lik
65{
65{
66 "lsp": {
66 "lsp": {
67 "go": {
67 "go": {
68− "disabled": false,
69 "command": "gopls"
68 "command": "gopls"
70 },
69 },
71 "typescript": {
70 "typescript": {
72− "disabled": false,
73 "command": "typescript-language-server",
71 "command": "typescript-language-server",
74 "args": ["--stdio"]
72 "args": ["--stdio"]
75 },
73 },
@@ -80,83 +78,30 @@ Crush can use LSPs for additional context to help inform its decisions, just lik
80}
78}
81```
79```
82
80
83−### Amazon Bedrock
84−
85−To use AWS Bedrock with Claude models, configure your AWS credentials and region:
86−
87−```json
88−{
89− "providers": [
90− {
91− "id": "bedrock",
92− "provider_type": "bedrock",
93− "extra_params": {
94− "region": "us-east-1"
95− }
96− }
97− ]
98−}
99−```
100−
101−Bedrock uses your AWS credentials from environment variables (`AWS_ACCESS_KEY_ID`, `AWS_SECRET_ACCESS_KEY`) or AWS credential profiles. The region can be specified in the configuration or via the `AWS_REGION` environment variable.
102−
103−### Google Vertex AI
104−
105−For Google Cloud Vertex AI with Gemini models, configure your project and location:
106−
107−```json
108−{
109− "providers": [
110− {
111− "id": "vertexai",
112− "provider_type": "vertexai",
113− "extra_headers": {
114− "project": "your-gcp-project-id",
115− "location": "us-central1"
116− }
117− }
118− ]
119−}
120−```
121−
122−Vertex AI uses Google Cloud authentication. Ensure you have the `GOOGLE_APPLICATION_CREDENTIALS` environment variable set or are authenticated via `gcloud auth application-default login`.
123−
124### OpenAI-Compatible APIs
81### OpenAI-Compatible APIs
125
82
126−Crush supports all OpenAI-compatible APIs, including local models via Ollama:
83+Crush supports all OpenAI-compatible APIs. Here's an example configuration for Deepseek, which uses an OpenAI-compatible API. Don't forget to set `DEEPSEEK_API_KEY` in your environment.
127
84
128```json
85```json
129{
86{
130− "providers": [
87+ "providers": {
131− {
88+ "deepseek": {
132− "id": "ollama",
133 "provider_type": "openai",
89 "provider_type": "openai",
134− "base_url": "http://localhost:11434/v1",
90+ "base_url": "https://api.deepseek.com/v1",
135 "models": [
91 "models": [
136 {
92 {
137− "id": "llama3.2:3b",
93+ "id": "deepseek-chat",
138− "name": "Llama 3.2 3B",
94+ "model": "Deepseek V3",
95+ "cost_per_1m_in": 0.27,
96+ "cost_per_1m_out": 1.1,
97+ "cost_per_1m_in_cached": 0.07,
98+ "cost_per_1m_out_cached": 1.1,
139− "context_window": 8192,
99+ "context_window": 64000,
140− "default_max_tokens": 4096
100+ "default_max_tokens": 5000
141 }
101 }
142 ]
102 ]
143 }
103 }
144− ]
145−}
104+ }
146−```
147−
148−For other OpenAI-compatible providers, adjust the `base_url` and provide an `api_key` if required:
149−
150−```json
151−{
152− "providers": [
153− {
154− "id": "custom-openai",
155− "provider_type": "openai",
156− "base_url": "https://api.example.com/v1",
157− "api_key": "your-api-key"
158− }
159− ]
160}
105}
161```
106```
162
107