on main help sign in
docs(readme): move "local models" section to be under "custom providers" (#1402)
025d604Andrey Nering
README.md+48โˆ’48
@@ -355,54 +355,6 @@ it creates. You can customize this behavior with the `attribution` option:
355- `co_authored_by`: When true (default), adds `Co-Authored-By: Crush <crush@charm.land>` to commit messages
355- `co_authored_by`: When true (default), adds `Co-Authored-By: Crush <crush@charm.land>` to commit messages
356- `generated_with`: When true (default), adds `๐Ÿ’˜ Generated with Crush` line to commit messages and PR descriptions
356- `generated_with`: When true (default), adds `๐Ÿ’˜ Generated with Crush` line to commit messages and PR descriptions
357
357
358โˆ’### Local Models
359โˆ’
360โˆ’Local models can also be configured via OpenAI-compatible API. Here are two common examples:
361โˆ’
362โˆ’#### Ollama
363โˆ’
364โˆ’```json
365โˆ’{
366โˆ’ "providers": {
367โˆ’ "ollama": {
368โˆ’ "name": "Ollama",
369โˆ’ "base_url": "http://localhost:11434/v1/",
370โˆ’ "type": "openai-compat",
371โˆ’ "models": [
372โˆ’ {
373โˆ’ "name": "Qwen 3 30B",
374โˆ’ "id": "qwen3:30b",
375โˆ’ "context_window": 256000,
376โˆ’ "default_max_tokens": 20000
377โˆ’ }
378โˆ’ ]
379โˆ’ }
380โˆ’ }
381โˆ’}
382โˆ’```
383โˆ’
384โˆ’#### LM Studio
385โˆ’
386โˆ’```json
387โˆ’{
388โˆ’ "providers": {
389โˆ’ "lmstudio": {
390โˆ’ "name": "LM Studio",
391โˆ’ "base_url": "http://localhost:1234/v1/",
392โˆ’ "type": "openai-compat",
393โˆ’ "models": [
394โˆ’ {
395โˆ’ "name": "Qwen 3 30B",
396โˆ’ "id": "qwen/qwen3-30b-a3b-2507",
397โˆ’ "context_window": 256000,
398โˆ’ "default_max_tokens": 20000
399โˆ’ }
400โˆ’ ]
401โˆ’ }
402โˆ’ }
403โˆ’}
404โˆ’```
405โˆ’
406### Custom Providers
358### Custom Providers
407
359
408Crush supports custom provider configurations for both OpenAI-compatible and
360Crush supports custom provider configurations for both OpenAI-compatible and
@@ -521,6 +473,54 @@ To add specific models to the configuration, configure as such:
521}
473}
522```
474```
523
475
476+### Local Models
477+
478+Local models can also be configured via OpenAI-compatible API. Here are two common examples:
479+
480+#### Ollama
481+
482+```json
483+{
484+ "providers": {
485+ "ollama": {
486+ "name": "Ollama",
487+ "base_url": "http://localhost:11434/v1/",
488+ "type": "openai-compat",
489+ "models": [
490+ {
491+ "name": "Qwen 3 30B",
492+ "id": "qwen3:30b",
493+ "context_window": 256000,
494+ "default_max_tokens": 20000
495+ }
496+ ]
497+ }
498+ }
499+}
500+```
501+
502+#### LM Studio
503+
504+```json
505+{
506+ "providers": {
507+ "lmstudio": {
508+ "name": "LM Studio",
509+ "base_url": "http://localhost:1234/v1/",
510+ "type": "openai-compat",
511+ "models": [
512+ {
513+ "name": "Qwen 3 30B",
514+ "id": "qwen/qwen3-30b-a3b-2507",
515+ "context_window": 256000,
516+ "default_max_tokens": 20000
517+ }
518+ ]
519+ }
520+ }
521+}
522+```
523+
524## Logging
524## Logging
525
525
526Sometimes you need to look at logs. Luckily, Crush logs all sorts of
526Sometimes you need to look at logs. Luckily, Crush logs all sorts of