mirror of
https://github.com/Aider-AI/aider.git
synced 2025-05-31 01:35:00 +00:00
Better unknown model warnings
This commit is contained in:
Paul Gauthier
parent
f1ce673f78
commit
efd3c39e50
4 changed files with 125 additions and 69 deletions
|
|
@ -137,10 +137,13 @@ class Coder:
|
|||
self.main_model = main_model
|
||||
|
||||
weak_model = main_model.weak_model
|
||||
self.io.tool_output(
|
||||
f"Models: {main_model.name} with {self.edit_format} edit format, weak model"
|
||||
f" {weak_model.name}"
|
||||
)
|
||||
prefix = "Model:"
|
||||
output = f" {main_model.name} with {self.edit_format} edit format"
|
||||
if weak_model is not main_model:
|
||||
prefix = "Models:"
|
||||
output += f", weak model {weak_model.name}"
|
||||
|
||||
self.io.tool_output(prefix + output)
|
||||
|
||||
self.show_diffs = show_diffs
|
||||
|
||||
|
|
|
|||
|
|
@ -21,6 +21,36 @@ os.environ["OR_SITE_URL"] = "http://aider.chat"
|
|||
os.environ["OR_APP_NAME"] = "Aider"
|
||||
|
||||
|
||||
def sanity_check_model(io, model):
|
||||
show = False
|
||||
|
||||
if model.missing_keys:
|
||||
show = True
|
||||
io.tool_error(f"Model {model}: Missing these environment variables:")
|
||||
for key in model.missing_keys:
|
||||
io.tool_error(f"- {key}")
|
||||
elif not model.keys_in_environment:
|
||||
show = True
|
||||
io.tool_error(f"Model {model}: Unknown which environment variables are required.")
|
||||
|
||||
if not model.info:
|
||||
show = True
|
||||
io.tool_error(
|
||||
f"Model {model}: Unknown model, context window size and token costs unavailable."
|
||||
)
|
||||
|
||||
possible_matches = models.fuzzy_match_models(model.name)
|
||||
if possible_matches:
|
||||
io.tool_error("Did you mean one of these?")
|
||||
for match in possible_matches:
|
||||
io.tool_error(f"- {match}")
|
||||
|
||||
if show:
|
||||
io.tool_error("For more info see https://aider.chat/docs/llms.html#model-warnings")
|
||||
|
||||
return False
|
||||
|
||||
|
||||
def get_git_root():
|
||||
"""Try and guess the git repo, since the conf.yml can be at the repo root"""
|
||||
try:
|
||||
|
|
@ -277,7 +307,7 @@ def main(argv=None, input=None, output=None, force_git_root=None):
|
|||
model_group.add_argument(
|
||||
"--require-model-info",
|
||||
action=argparse.BooleanOptionalAction,
|
||||
default=True,
|
||||
default=False,
|
||||
help="Only work with models that have meta-data available (default: True)",
|
||||
)
|
||||
model_group.add_argument(
|
||||
|
|
@ -602,13 +632,15 @@ def main(argv=None, input=None, output=None, force_git_root=None):
|
|||
if args.openai_organization_id:
|
||||
os.environ["OPENAI_ORGANIZATION"] = args.openai_organization_id
|
||||
|
||||
# Check in advance that we have model metadata
|
||||
try:
|
||||
main_model = models.Model(
|
||||
args.model, weak_model=args.weak_model, require_model_info=args.require_model_info
|
||||
)
|
||||
except (models.NoModelInfo, models.ModelEnvironmentError) as err:
|
||||
io.tool_error(str(err))
|
||||
main_model = models.Model(args.model, weak_model=args.weak_model)
|
||||
|
||||
missing_model_info = False
|
||||
if not sanity_check_model(io, main_model):
|
||||
missing_model_info = True
|
||||
if main_model.weak_model and main_model.weak_model is not main_model:
|
||||
if not sanity_check_model(io, main_model.weak_model):
|
||||
missing_model_info = True
|
||||
if args.require_model_info and missing_model_info:
|
||||
return 1
|
||||
|
||||
try:
|
||||
|
|
|
|||
|
|
@ -13,24 +13,6 @@ from aider.dump import dump # noqa: F401
|
|||
DEFAULT_MODEL_NAME = "gpt-4-1106-preview"
|
||||
|
||||
|
||||
class NoModelInfo(Exception):
|
||||
"""
|
||||
Exception raised when model information cannot be retrieved.
|
||||
"""
|
||||
|
||||
def __init__(self, model):
|
||||
super().__init__(check_model_name(model))
|
||||
|
||||
|
||||
class ModelEnvironmentError(Exception):
|
||||
"""
|
||||
Exception raised when the environment isn't setup for the model
|
||||
"""
|
||||
|
||||
def __init__(self, message):
|
||||
super().__init__(message)
|
||||
|
||||
|
||||
@dataclass
|
||||
class ModelSettings:
|
||||
name: str
|
||||
|
|
@ -164,30 +146,18 @@ class Model:
|
|||
max_chat_history_tokens = 1024
|
||||
weak_model = None
|
||||
|
||||
def __init__(self, model, weak_model=None, require_model_info=True, validate_environment=True):
|
||||
def __init__(self, model, weak_model=None):
|
||||
self.name = model
|
||||
|
||||
# Are all needed keys/params available?
|
||||
res = litellm.validate_environment(model)
|
||||
missing_keys = res.get("missing_keys")
|
||||
keys_in_environment = res.get("keys_in_environment")
|
||||
|
||||
if missing_keys:
|
||||
if validate_environment:
|
||||
res = f"To use model {model}, please set these environment variables:"
|
||||
for key in missing_keys:
|
||||
res += f"- {key}"
|
||||
raise ModelEnvironmentError(res)
|
||||
elif not keys_in_environment:
|
||||
# https://github.com/BerriAI/litellm/issues/3190
|
||||
print(f"Unable to check environment variables for model {model}")
|
||||
self.missing_keys = res.get("missing_keys")
|
||||
self.keys_in_environment = res.get("keys_in_environment")
|
||||
|
||||
# Do we have the model_info?
|
||||
try:
|
||||
self.info = litellm.get_model_info(model)
|
||||
except Exception:
|
||||
if require_model_info:
|
||||
raise NoModelInfo(model)
|
||||
self.info = dict()
|
||||
|
||||
if self.info.get("max_input_tokens", 0) < 32 * 1024:
|
||||
|
|
@ -199,7 +169,7 @@ class Model:
|
|||
if weak_model is False:
|
||||
self.weak_model_name = None
|
||||
else:
|
||||
self.get_weak_model(weak_model, require_model_info)
|
||||
self.get_weak_model(weak_model)
|
||||
|
||||
def configure_model_settings(self, model):
|
||||
for ms in MODEL_SETTINGS:
|
||||
|
|
@ -210,7 +180,9 @@ class Model:
|
|||
setattr(self, field.name, val)
|
||||
return # <--
|
||||
|
||||
if "llama3" in model and "70b" in model:
|
||||
model = model.lower()
|
||||
|
||||
if ("llama3" in model or "llama-3" in model) and "70b" in model:
|
||||
self.edit_format = "diff"
|
||||
self.use_repo_map = True
|
||||
self.send_undo_reply = True
|
||||
|
|
@ -235,7 +207,7 @@ class Model:
|
|||
def __str__(self):
|
||||
return self.name
|
||||
|
||||
def get_weak_model(self, provided_weak_model_name, require_model_info):
|
||||
def get_weak_model(self, provided_weak_model_name):
|
||||
# If weak_model_name is provided, override the model settings
|
||||
if provided_weak_model_name:
|
||||
self.weak_model_name = provided_weak_model_name
|
||||
|
|
@ -251,7 +223,6 @@ class Model:
|
|||
self.weak_model = Model(
|
||||
self.weak_model_name,
|
||||
weak_model=False,
|
||||
require_model_info=require_model_info,
|
||||
)
|
||||
return self.weak_model
|
||||
|
||||
|
|
@ -313,19 +284,6 @@ class Model:
|
|||
return img.size
|
||||
|
||||
|
||||
def check_model_name(model):
|
||||
res = f"Unknown model {model}"
|
||||
|
||||
possible_matches = fuzzy_match_models(model)
|
||||
|
||||
if possible_matches:
|
||||
res += ", did you mean one of these?"
|
||||
for match in possible_matches:
|
||||
res += "\n- " + match
|
||||
|
||||
return res
|
||||
|
||||
|
||||
def fuzzy_match_models(name):
|
||||
models = litellm.model_cost.keys()
|
||||
|
||||
|
|
|
|||
77
docs/llms.md
77
docs/llms.md
|
|
@ -29,6 +29,7 @@ So you should expect that models which are less capable than GPT-3.5 may struggl
|
|||
- [OpenRouter](#openrouter)
|
||||
- [OpenAI compatible APIs](#openai-compatible-apis)
|
||||
- [Other LLMs](#other-llms)
|
||||
- [Model warnings](#model-warnings)
|
||||
- [Editing format](#editing-format)
|
||||
|
||||
## OpenAI
|
||||
|
|
@ -136,15 +137,19 @@ aider --model openrouter/<provider>/<model>
|
|||
|
||||
## OpenAI compatible APIs
|
||||
|
||||
If your LLM is accessible via an OpenAI compatible API endpoint,
|
||||
you can use `--openai-api-base` to have aider connect to it.
|
||||
Aider can connect to any LLM which is accessible via an OpenAI compatible API endpoint.
|
||||
Use `--openai-api-base` or set the `OPENAI_API_BASE`
|
||||
environment variable to have aider connect to it.
|
||||
|
||||
You might need to use `--no-require-model-info` if aider doesn't
|
||||
recognize the model you want to use.
|
||||
For unknown models, aider won't have normal metadata available like
|
||||
the context window size, token costs, etc.
|
||||
Some minor functionality will be limited when using such models.
|
||||
```
|
||||
export OPENAI_API_BASE=<your-endpoint-goes-here>
|
||||
export OPENAI_API_KEY=<your-key-goes-here-if-required>
|
||||
aider --model <model-name>
|
||||
```
|
||||
|
||||
See the [Model warnings](#model-warnings)
|
||||
section for information on warnings which will occur
|
||||
when working with models that aider is not familiar with.
|
||||
|
||||
## Other LLMs
|
||||
|
||||
|
|
@ -178,6 +183,64 @@ error message listing which parameters are needed.
|
|||
See the [list of providers supported by litellm](https://docs.litellm.ai/docs/providers)
|
||||
for more details.
|
||||
|
||||
## Model warnings
|
||||
|
||||
On startup, aider tries to sanity check that it is configured correctly
|
||||
to work with the specified models:
|
||||
|
||||
- It checks to see that all required environment variables are set for the model. These variables are required to configure things like API keys, API base URLs, etc.
|
||||
- It checks a metadata database to look up the context window size and token costs for the model.
|
||||
|
||||
Sometimes one or both of these checks will fail, so aider will issue
|
||||
some of the following warnings.
|
||||
|
||||
##### Missing environment variables
|
||||
|
||||
You need to set the listed variables.
|
||||
Otherwise you will get error messages when you attempt to chat with the model.
|
||||
|
||||
```
|
||||
Model azure/gpt-4-turbo: Missing these environment variables:
|
||||
- AZURE_API_BASE
|
||||
- AZURE_API_VERSION
|
||||
- AZURE_API_KEY
|
||||
```
|
||||
|
||||
##### Unknown which environment variables are required
|
||||
|
||||
Aider is unable verify the environment because it doesn't know
|
||||
which variables are required for the model.
|
||||
If required variables are missing,
|
||||
you may get errors when you attempt to chat with the model.
|
||||
You can look in the
|
||||
[litellm provider documentation](https://docs.litellm.ai/docs/providers)
|
||||
to see if the required variables are listed there.
|
||||
```
|
||||
Model gpt-5: Unknown which environment variables are required.
|
||||
```
|
||||
|
||||
|
||||
##### Unknown model, did you mean?
|
||||
|
||||
If you specify a model that aider has never heard of, you will get an
|
||||
"unknown model" warning.
|
||||
|
||||
In this case, aider won't have normal metadata available like
|
||||
the context window size, token costs, etc.
|
||||
Some minor functionality will be limited when using such models, but
|
||||
it's not really a significant problem.
|
||||
|
||||
Aider will also try to suggest similarly named models,
|
||||
in case you made a typo or mistake when specifying the model name.
|
||||
|
||||
```
|
||||
Model gpt-5: Unknown model, context window size and token costs unavailable.
|
||||
Did you mean one of these?
|
||||
- gpt-4
|
||||
```
|
||||
|
||||
|
||||
|
||||
|
||||
## Editing format
|
||||
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue