luca-litellm 0.1.0__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -0,0 +1,201 @@
1
+ Apache License
2
+ Version 2.0, January 2004
3
+ http://www.apache.org/licenses/
4
+
5
+ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
6
+
7
+ 1. Definitions.
8
+
9
+ "License" shall mean the terms and conditions for use, reproduction,
10
+ and distribution as defined by Sections 1 through 9 of this document.
11
+
12
+ "Licensor" shall mean the copyright owner or entity authorized by
13
+ the copyright owner that is granting the License.
14
+
15
+ "Legal Entity" shall mean the union of the acting entity and all
16
+ other entities that control, are controlled by, or are under common
17
+ control with that entity. For the purposes of this definition,
18
+ "control" means (i) the power, direct or indirect, to cause the
19
+ direction or management of such entity, whether by contract or
20
+ otherwise, or (ii) ownership of fifty percent (50%) or more of the
21
+ outstanding shares, or (iii) beneficial ownership of such entity.
22
+
23
+ "You" (or "Your") shall mean an individual or Legal Entity
24
+ exercising permissions granted by this License.
25
+
26
+ "Source" form shall mean the preferred form for making modifications,
27
+ including but not limited to software source code, documentation
28
+ source, and configuration files.
29
+
30
+ "Object" form shall mean any form resulting from mechanical
31
+ transformation or translation of a Source form, including but
32
+ not limited to compiled object code, generated documentation,
33
+ and conversions to other media types.
34
+
35
+ "Work" shall mean the work of authorship, whether in Source or
36
+ Object form, made available under the License, as indicated by a
37
+ copyright notice that is included in or attached to the work
38
+ (an example is provided in the Appendix below).
39
+
40
+ "Derivative Works" shall mean any work, whether in Source or Object
41
+ form, that is based on (or derived from) the Work and for which the
42
+ editorial revisions, annotations, elaborations, or other modifications
43
+ represent, as a whole, an original work of authorship. For the purposes
44
+ of this License, Derivative Works shall not include works that remain
45
+ separable from, or merely link (or bind by name) to the interfaces of,
46
+ the Work and Derivative Works thereof.
47
+
48
+ "Contribution" shall mean any work of authorship, including
49
+ the original version of the Work and any modifications or additions
50
+ to that Work or Derivative Works thereof, that is intentionally
51
+ submitted to Licensor for inclusion in the Work by the copyright owner
52
+ or by an individual or Legal Entity authorized to submit on behalf of
53
+ the copyright owner. For the purposes of this definition, "submitted"
54
+ means any form of electronic, verbal, or written communication sent
55
+ to the Licensor or its representatives, including but not limited to
56
+ communication on electronic mailing lists, source code control systems,
57
+ and issue tracking systems that are managed by, or on behalf of, the
58
+ Licensor for the purpose of discussing and improving the Work, but
59
+ excluding communication that is conspicuously marked or otherwise
60
+ designated in writing by the copyright owner as "Not a Contribution."
61
+
62
+ "Contributor" shall mean Licensor and any individual or Legal Entity
63
+ on behalf of whom a Contribution has been received by Licensor and
64
+ subsequently incorporated within the Work.
65
+
66
+ 2. Grant of Copyright License. Subject to the terms and conditions of
67
+ this License, each Contributor hereby grants to You a perpetual,
68
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
69
+ copyright license to reproduce, prepare Derivative Works of,
70
+ publicly display, publicly perform, sublicense, and distribute the
71
+ Work and such Derivative Works in Source or Object form.
72
+
73
+ 3. Grant of Patent License. Subject to the terms and conditions of
74
+ this License, each Contributor hereby grants to You a perpetual,
75
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
76
+ (except as stated in this section) patent license to make, have made,
77
+ use, offer to sell, sell, import, and otherwise transfer the Work,
78
+ where such license applies only to those patent claims licensable
79
+ by such Contributor that are necessarily infringed by their
80
+ Contribution(s) alone or by combination of their Contribution(s)
81
+ with the Work to which such Contribution(s) was submitted. If You
82
+ institute patent litigation against any entity (including a
83
+ cross-claim or counterclaim in a lawsuit) alleging that the Work
84
+ or a Contribution incorporated within the Work constitutes direct
85
+ or contributory patent infringement, then any patent licenses
86
+ granted to You under this License for that Work shall terminate
87
+ as of the date such litigation is filed.
88
+
89
+ 4. Redistribution. You may reproduce and distribute copies of the
90
+ Work or Derivative Works thereof in any medium, with or without
91
+ modifications, and in Source or Object form, provided that You
92
+ meet the following conditions:
93
+
94
+ (a) You must give any other recipients of the Work or
95
+ Derivative Works a copy of this License; and
96
+
97
+ (b) You must cause any modified files to carry prominent notices
98
+ stating that You changed the files; and
99
+
100
+ (c) You must retain, in the Source form of any Derivative Works
101
+ that You distribute, all copyright, patent, trademark, and
102
+ attribution notices from the Source form of the Work,
103
+ excluding those notices that do not pertain to any part of
104
+ the Derivative Works; and
105
+
106
+ (d) If the Work includes a "NOTICE" text file as part of its
107
+ distribution, then any Derivative Works that You distribute must
108
+ include a readable copy of the attribution notices contained
109
+ within such NOTICE file, excluding those notices that do not
110
+ pertain to any part of the Derivative Works, in at least one
111
+ of the following places: within a NOTICE text file distributed
112
+ as part of the Derivative Works; within the Source form or
113
+ documentation, if provided along with the Derivative Works; or,
114
+ within a display generated by the Derivative Works, if and
115
+ wherever such third-party notices normally appear. The contents
116
+ of the NOTICE file are for informational purposes only and
117
+ do not modify the License. You may add Your own attribution
118
+ notices within Derivative Works that You distribute, alongside
119
+ or as an addendum to the NOTICE text from the Work, provided
120
+ that such additional attribution notices cannot be construed
121
+ as modifying the License.
122
+
123
+ You may add Your own copyright statement to Your modifications and
124
+ may provide additional or different license terms and conditions
125
+ for use, reproduction, or distribution of Your modifications, or
126
+ for any such Derivative Works as a whole, provided Your use,
127
+ reproduction, and distribution of the Work otherwise complies with
128
+ the conditions stated in this License.
129
+
130
+ 5. Submission of Contributions. Unless You explicitly state otherwise,
131
+ any Contribution intentionally submitted for inclusion in the Work
132
+ by You to the Licensor shall be under the terms and conditions of
133
+ this License, without any additional terms or conditions.
134
+ Notwithstanding the above, nothing herein shall supersede or modify
135
+ the terms of any separate license agreement you may have executed
136
+ with Licensor regarding such Contributions.
137
+
138
+ 6. Trademarks. This License does not grant permission to use the trade
139
+ names, trademarks, service marks, or product names of the Licensor,
140
+ except as required for reasonable and customary use in describing the
141
+ origin of the Work and reproducing the content of the NOTICE file.
142
+
143
+ 7. Disclaimer of Warranty. Unless required by applicable law or
144
+ agreed to in writing, Licensor provides the Work (and each
145
+ Contributor provides its Contributions) on an "AS IS" BASIS,
146
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
147
+ implied, including, without limitation, any warranties or conditions
148
+ of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
149
+ PARTICULAR PURPOSE. You are solely responsible for determining the
150
+ appropriateness of using or redistributing the Work and assume any
151
+ risks associated with Your exercise of permissions under this License.
152
+
153
+ 8. Limitation of Liability. In no event and under no legal theory,
154
+ whether in tort (including negligence), contract, or otherwise,
155
+ unless required by applicable law (such as deliberate and grossly
156
+ negligent acts) or agreed to in writing, shall any Contributor be
157
+ liable to You for damages, including any direct, indirect, special,
158
+ incidental, or consequential damages of any character arising as a
159
+ result of this License or out of the use or inability to use the
160
+ Work (including but not limited to damages for loss of goodwill,
161
+ work stoppage, computer failure or malfunction, or any and all
162
+ other commercial damages or losses), even if such Contributor
163
+ has been advised of the possibility of such damages.
164
+
165
+ 9. Accepting Warranty or Additional Liability. While redistributing
166
+ the Work or Derivative Works thereof, You may choose to offer,
167
+ and charge a fee for, acceptance of support, warranty, indemnity,
168
+ or other liability obligations and/or rights consistent with this
169
+ License. However, in accepting such obligations, You may act only
170
+ on Your own behalf and on Your sole responsibility, not on behalf
171
+ of any other Contributor, and only if You agree to indemnify,
172
+ defend, and hold each Contributor harmless for any liability
173
+ incurred by, or claims asserted against, such Contributor by reason
174
+ of your accepting any such warranty or additional liability.
175
+
176
+ END OF TERMS AND CONDITIONS
177
+
178
+ APPENDIX: How to apply the Apache License to your work.
179
+
180
+ To apply the Apache License to your work, attach the following
181
+ boilerplate notice, with the fields enclosed by brackets "{}"
182
+ replaced with your own identifying information. (Don't include
183
+ the brackets!) The text should be enclosed in the appropriate
184
+ comment syntax for the file format. We also recommend that a
185
+ file or class name and description of purpose be included on the
186
+ same "printed page" as the copyright notice for easier
187
+ identification within third-party archives.
188
+
189
+ Copyright 2026 Luca
190
+
191
+ Licensed under the Apache License, Version 2.0 (the "License");
192
+ you may not use this file except in compliance with the License.
193
+ You may obtain a copy of the License at
194
+
195
+ http://www.apache.org/licenses/LICENSE-2.0
196
+
197
+ Unless required by applicable law or agreed to in writing, software
198
+ distributed under the License is distributed on an "AS IS" BASIS,
199
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
200
+ See the License for the specific language governing permissions and
201
+ limitations under the License.
@@ -0,0 +1,4 @@
1
+ luca-litellm
2
+ Copyright 2026 Luca
3
+
4
+ This product is licensed under the Apache License, Version 2.0 (LICENSE).
@@ -0,0 +1,74 @@
1
+ Metadata-Version: 2.4
2
+ Name: luca-litellm
3
+ Version: 0.1.0
4
+ Summary: Send the cost of each LiteLLM call to Luca for accounting, without prompts or responses
5
+ Author-email: Luca <hello@luca.network>
6
+ License: Apache-2.0
7
+ Project-URL: Homepage, https://luca.network/developers
8
+ Project-URL: Documentation, https://luca.network/developers#litellm
9
+ Project-URL: Repository, https://github.com/luca-network/luca-litellm
10
+ Keywords: litellm,ai spend,accounting,finops,cost of revenue,llm cost,gateway
11
+ Classifier: License :: OSI Approved :: Apache Software License
12
+ Classifier: Programming Language :: Python :: 3
13
+ Classifier: Topic :: Office/Business :: Financial :: Accounting
14
+ Requires-Python: >=3.9
15
+ Description-Content-Type: text/markdown
16
+ License-File: LICENSE
17
+ License-File: NOTICE
18
+ Requires-Dist: litellm>=1.70
19
+ Requires-Dist: httpx>=0.27
20
+ Dynamic: license-file
21
+
22
+ # luca-litellm
23
+
24
+ Send the cost of every call through your LiteLLM proxy to Luca, so your finance team gets AI spend by customer, team and model, accrued and matched to each provider's invoice, with journal entries for the ERP.
25
+
26
+ Only the fields accounting needs leave your network: when the call was made, the provider and model, the cost and token counts, and the team, key, project, customer, tags and MCP server LiteLLM records with it. The plugin drops the prompt, the response, tool arguments and tool results before anything is sent. Email addresses are left out unless you turn them on.
27
+
28
+ ## Set up
29
+
30
+ 1. In Luca, add a **LiteLLM AI gateway** connection. Luca shows its webhook URL. Make an ingest token for it:
31
+ `openssl rand -hex 32`
32
+ 2. On the machine that runs the proxy:
33
+ ```
34
+ pip install luca-litellm
35
+ ```
36
+ 3. Add the plugin to the proxy's `config.yaml`:
37
+ ```yaml
38
+ litellm_settings:
39
+ callbacks: ["luca_litellm.handler"]
40
+ ```
41
+ 4. Set two environment variables for the proxy:
42
+ ```
43
+ LUCA_LITELLM_URL=<the webhook URL Luca shows>
44
+ LUCA_LITELLM_TOKEN=<the ingest token>
45
+ ```
46
+ 5. Restart the proxy.
47
+
48
+ The plugin sends calls in batches (every 10 seconds, or every 200 calls), outside the request path, so it adds no time to your calls. If Luca can't be reached, it keeps the calls and tries again.
49
+
50
+ ## Which customer a call served
51
+
52
+ LiteLLM records a customer for each call when your app passes `user` in the request, or when the call uses a key or team set up for that customer. Luca uses it to show AI cost by customer.
53
+
54
+ ## Options
55
+
56
+ | Variable | What it does |
57
+ |---|---|
58
+ | `LUCA_LITELLM_URL` | The connection's webhook URL (required) |
59
+ | `LUCA_LITELLM_TOKEN` | The connection's ingest token (required) |
60
+ | `LUCA_LITELLM_INCLUDE_EMAIL` | `true` to send the user's email address with each call (off by default) |
61
+
62
+ To use it from the LiteLLM Python SDK instead of the proxy:
63
+
64
+ ```python
65
+ import litellm
66
+ from luca_litellm import LucaLogger
67
+ litellm.callbacks = [LucaLogger(url="https://…", token="…")]
68
+ ```
69
+
70
+ ## Without installing a package
71
+
72
+ LiteLLM's built-in webhook callback can send the same calls. Luca shows that configuration on the connection too. With it, turn message logging off (`turn_off_message_logging: true`) so prompts and responses are not sent.
73
+
74
+ Apache-2.0. Questions: hello@luca.network
@@ -0,0 +1,53 @@
1
+ # luca-litellm
2
+
3
+ Send the cost of every call through your LiteLLM proxy to Luca, so your finance team gets AI spend by customer, team and model, accrued and matched to each provider's invoice, with journal entries for the ERP.
4
+
5
+ Only the fields accounting needs leave your network: when the call was made, the provider and model, the cost and token counts, and the team, key, project, customer, tags and MCP server LiteLLM records with it. The plugin drops the prompt, the response, tool arguments and tool results before anything is sent. Email addresses are left out unless you turn them on.
6
+
7
+ ## Set up
8
+
9
+ 1. In Luca, add a **LiteLLM AI gateway** connection. Luca shows its webhook URL. Make an ingest token for it:
10
+ `openssl rand -hex 32`
11
+ 2. On the machine that runs the proxy:
12
+ ```
13
+ pip install luca-litellm
14
+ ```
15
+ 3. Add the plugin to the proxy's `config.yaml`:
16
+ ```yaml
17
+ litellm_settings:
18
+ callbacks: ["luca_litellm.handler"]
19
+ ```
20
+ 4. Set two environment variables for the proxy:
21
+ ```
22
+ LUCA_LITELLM_URL=<the webhook URL Luca shows>
23
+ LUCA_LITELLM_TOKEN=<the ingest token>
24
+ ```
25
+ 5. Restart the proxy.
26
+
27
+ The plugin sends calls in batches (every 10 seconds, or every 200 calls), outside the request path, so it adds no time to your calls. If Luca can't be reached, it keeps the calls and tries again.
28
+
29
+ ## Which customer a call served
30
+
31
+ LiteLLM records a customer for each call when your app passes `user` in the request, or when the call uses a key or team set up for that customer. Luca uses it to show AI cost by customer.
32
+
33
+ ## Options
34
+
35
+ | Variable | What it does |
36
+ |---|---|
37
+ | `LUCA_LITELLM_URL` | The connection's webhook URL (required) |
38
+ | `LUCA_LITELLM_TOKEN` | The connection's ingest token (required) |
39
+ | `LUCA_LITELLM_INCLUDE_EMAIL` | `true` to send the user's email address with each call (off by default) |
40
+
41
+ To use it from the LiteLLM Python SDK instead of the proxy:
42
+
43
+ ```python
44
+ import litellm
45
+ from luca_litellm import LucaLogger
46
+ litellm.callbacks = [LucaLogger(url="https://…", token="…")]
47
+ ```
48
+
49
+ ## Without installing a package
50
+
51
+ LiteLLM's built-in webhook callback can send the same calls. Luca shows that configuration on the connection too. With it, turn message logging off (`turn_off_message_logging: true`) so prompts and responses are not sent.
52
+
53
+ Apache-2.0. Questions: hello@luca.network
@@ -0,0 +1,241 @@
1
+ """Send the cost of each LiteLLM call to Luca, without prompts or responses.
2
+
3
+ Add to the proxy's config.yaml:
4
+
5
+ litellm_settings:
6
+ callbacks: ["luca_litellm.handler"]
7
+
8
+ and set two environment variables on the proxy:
9
+
10
+ LUCA_LITELLM_URL the webhook URL Luca shows for the connection
11
+ LUCA_LITELLM_TOKEN the ingest token you gave the connection
12
+
13
+ Only the fields accounting needs leave the proxy: when the call was made, the provider and model, the cost and
14
+ token counts, and the team, key, project, customer, tags and MCP server LiteLLM records with it. The prompt, the
15
+ response, tool arguments and tool results are dropped here, before anything is sent.
16
+ """
17
+
18
+ from __future__ import annotations
19
+
20
+ import asyncio
21
+ import atexit
22
+ import logging
23
+ import os
24
+ import threading
25
+ from typing import Any
26
+ from urllib.parse import urlsplit
27
+
28
+ import httpx
29
+ from litellm.integrations.custom_logger import CustomLogger
30
+
31
+ __all__ = ["LucaLogger", "to_record", "handler", "__version__"]
32
+ __version__ = "0.1.0"
33
+
34
+ log = logging.getLogger("luca_litellm")
35
+
36
+ # The top-level fields of LiteLLM's standard logging payload that Luca reads. Nothing else is sent.
37
+ _TOP = (
38
+ "id", "status", "call_type", "startTime", "endTime", "model", "model_group", "model_id", "custom_llm_provider",
39
+ "response_cost", "prompt_tokens", "completion_tokens", "total_tokens", "cache_hit", "end_user",
40
+ )
41
+ # From the call's metadata: who made it and for whom. The user's email address is left out unless
42
+ # LUCA_LITELLM_INCLUDE_EMAIL is set; the user's ID is enough for Luca to split usage by user.
43
+ _META = (
44
+ "user_api_key_hash", "user_api_key_alias", "user_api_key_team_id", "user_api_key_team_alias",
45
+ "user_api_key_user_id", "user_api_key_project_id", "user_api_key_project_alias", "user_api_key_end_user_id",
46
+ "user_api_key_org_id", "user_api_key_org_alias", "team_id", "team_alias",
47
+ )
48
+ # From the usage object: only the cached-token counts, which change the cost.
49
+ _USAGE = ("cache_read_input_tokens", "cache_creation_input_tokens")
50
+
51
+ _MAX_QUEUE = 20_000
52
+
53
+
54
+ def _host(api_base: Any) -> str | None:
55
+ """The provider's address without its path or query, which can hold a deployment name or a key."""
56
+ if not isinstance(api_base, str) or not api_base:
57
+ return None
58
+ p = urlsplit(api_base if "://" in api_base else f"https://{api_base}")
59
+ return f"{p.scheme}://{p.hostname}" if p.hostname else None
60
+
61
+
62
+ def to_record(payload: Any, include_email: bool = False) -> dict | None:
63
+ """The part of one call that Luca keeps, or None when there is nothing to send (no payload, a failed call)."""
64
+ if not isinstance(payload, dict) or not payload.get("id"):
65
+ return None
66
+ if payload.get("status") not in (None, "success"):
67
+ return None
68
+ rec: dict[str, Any] = {k: payload[k] for k in _TOP if payload.get(k) is not None}
69
+ host = _host(payload.get("api_base"))
70
+ if host:
71
+ rec["api_base"] = host
72
+ tags = payload.get("request_tags")
73
+ # LiteLLM adds the client's User-Agent as tags; those say nothing about who the call served.
74
+ tags = [str(t) for t in tags if not str(t).startswith("User-Agent:")] if isinstance(tags, list) else []
75
+ if tags:
76
+ rec["request_tags"] = tags
77
+ md = payload.get("metadata") or {}
78
+ meta: dict[str, Any] = {k: md[k] for k in _META if md.get(k) is not None}
79
+ if include_email and md.get("user_api_key_user_email"):
80
+ meta["user_api_key_user_email"] = md["user_api_key_user_email"]
81
+ usage = md.get("usage_object") or {}
82
+ if isinstance(usage, dict):
83
+ u: dict[str, Any] = {k: usage[k] for k in _USAGE if usage.get(k) is not None}
84
+ cached = (usage.get("prompt_tokens_details") or {}).get("cached_tokens")
85
+ if cached is not None:
86
+ u["prompt_tokens_details"] = {"cached_tokens": cached}
87
+ if u:
88
+ meta["usage_object"] = u
89
+ mcp = md.get("mcp_tool_call_metadata") or {}
90
+ if isinstance(mcp, dict) and (mcp.get("mcp_server_name") or mcp.get("name")):
91
+ meta["mcp_tool_call_metadata"] = {k: mcp[k] for k in ("mcp_server_name", "name") if mcp.get(k)}
92
+ if meta:
93
+ rec["metadata"] = meta
94
+ return rec
95
+
96
+
97
+ class LucaLogger(CustomLogger):
98
+ """A LiteLLM callback that batches each successful call's cost fields and posts them to Luca."""
99
+
100
+ def __init__(self, url: str | None = None, token: str | None = None, batch_size: int = 200,
101
+ flush_seconds: float = 10.0, timeout: float = 10.0, include_email: bool | None = None, **kwargs: Any):
102
+ super().__init__(**kwargs)
103
+ self._url = url
104
+ self._token = token
105
+ self.batch_size = max(1, min(int(batch_size), 1000))
106
+ self.flush_seconds = max(1.0, float(flush_seconds))
107
+ self.timeout = float(timeout)
108
+ self._include_email = include_email
109
+ self._queue: list[dict] = []
110
+ self._lock = threading.Lock()
111
+ self._task: asyncio.Task | None = None
112
+ self._warned = False
113
+ self._seen: dict[str, None] = {}
114
+ atexit.register(self.flush)
115
+
116
+ # Settings are read when first needed, so the module can be imported before the environment is set.
117
+ @property
118
+ def url(self) -> str | None:
119
+ return self._url or os.environ.get("LUCA_LITELLM_URL")
120
+
121
+ @property
122
+ def token(self) -> str | None:
123
+ return self._token or os.environ.get("LUCA_LITELLM_TOKEN")
124
+
125
+ @property
126
+ def include_email(self) -> bool:
127
+ if self._include_email is not None:
128
+ return self._include_email
129
+ return os.environ.get("LUCA_LITELLM_INCLUDE_EMAIL", "").lower() in ("1", "true", "yes")
130
+
131
+ def _ready(self) -> bool:
132
+ if self.url and self.token:
133
+ return True
134
+ if not self._warned:
135
+ log.warning("luca_litellm: set LUCA_LITELLM_URL and LUCA_LITELLM_TOKEN; nothing is sent until both are set")
136
+ self._warned = True
137
+ return False
138
+
139
+ def _add(self, kwargs: dict) -> bool:
140
+ rec = to_record(kwargs.get("standard_logging_object"), self.include_email)
141
+ if rec is None:
142
+ return False
143
+ with self._lock:
144
+ # LiteLLM can report one call to both the sync and the async hook; send it once.
145
+ if rec["id"] in self._seen:
146
+ return False
147
+ self._seen[rec["id"]] = None
148
+ if len(self._seen) > _MAX_QUEUE:
149
+ del self._seen[next(iter(self._seen))]
150
+ self._queue.append(rec)
151
+ if len(self._queue) > _MAX_QUEUE:
152
+ dropped = len(self._queue) - _MAX_QUEUE
153
+ del self._queue[:dropped]
154
+ log.warning("luca_litellm: Luca could not be reached; %d oldest calls dropped from the queue", dropped)
155
+ return len(self._queue) >= self.batch_size
156
+
157
+ def _take(self) -> list[dict]:
158
+ with self._lock:
159
+ batch, self._queue = self._queue[: self.batch_size], self._queue[self.batch_size:]
160
+ return batch
161
+
162
+ def _put_back(self, batch: list[dict]) -> None:
163
+ with self._lock:
164
+ self._queue[:0] = batch
165
+
166
+ def _headers(self) -> dict[str, str]:
167
+ return {"Authorization": f"Bearer {self.token}", "Content-Type": "application/json",
168
+ "User-Agent": f"luca-litellm/{__version__}"}
169
+
170
+ @staticmethod
171
+ def _ok(res: httpx.Response) -> bool:
172
+ if res.status_code < 300:
173
+ return True
174
+ # A rejected delivery (bad token, wrong URL) would fail again; keep it only for errors that may pass.
175
+ if res.status_code in (408, 429) or res.status_code >= 500:
176
+ return False
177
+ log.error("luca_litellm: Luca refused the delivery (%s): %s", res.status_code, res.text[:300])
178
+ return True
179
+
180
+ async def async_flush(self) -> None:
181
+ if not self._ready():
182
+ return
183
+ async with httpx.AsyncClient(timeout=self.timeout) as client:
184
+ while True:
185
+ batch = self._take()
186
+ if not batch:
187
+ return
188
+ try:
189
+ if not self._ok(await client.post(self.url, json=batch, headers=self._headers())):
190
+ self._put_back(batch)
191
+ return
192
+ except httpx.HTTPError as e:
193
+ log.warning("luca_litellm: could not reach Luca (%s); will retry", e)
194
+ self._put_back(batch)
195
+ return
196
+
197
+ def flush(self) -> None:
198
+ if not self._queue or not self._ready():
199
+ return
200
+ with httpx.Client(timeout=self.timeout) as client:
201
+ while True:
202
+ batch = self._take()
203
+ if not batch:
204
+ return
205
+ try:
206
+ if not self._ok(client.post(self.url, json=batch, headers=self._headers())):
207
+ self._put_back(batch)
208
+ return
209
+ except httpx.HTTPError as e:
210
+ log.warning("luca_litellm: could not reach Luca (%s); will retry", e)
211
+ self._put_back(batch)
212
+ return
213
+
214
+ async def _every(self) -> None:
215
+ while True:
216
+ await asyncio.sleep(self.flush_seconds)
217
+ try:
218
+ await self.async_flush()
219
+ except Exception: # never let the background loop die
220
+ log.exception("luca_litellm: flush failed")
221
+
222
+ async def async_log_success_event(self, kwargs, response_obj, start_time, end_time):
223
+ try:
224
+ full = self._add(kwargs)
225
+ if self._task is None or self._task.done():
226
+ self._task = asyncio.get_running_loop().create_task(self._every())
227
+ if full:
228
+ await self.async_flush()
229
+ except Exception: # logging must never affect the call itself
230
+ log.exception("luca_litellm: could not record a call")
231
+
232
+ def log_success_event(self, kwargs, response_obj, start_time, end_time):
233
+ try:
234
+ if self._add(kwargs):
235
+ self.flush()
236
+ except Exception:
237
+ log.exception("luca_litellm: could not record a call")
238
+
239
+
240
+ # The instance the proxy's config names: callbacks: ["luca_litellm.handler"]
241
+ handler = LucaLogger()
@@ -0,0 +1,74 @@
1
+ Metadata-Version: 2.4
2
+ Name: luca-litellm
3
+ Version: 0.1.0
4
+ Summary: Send the cost of each LiteLLM call to Luca for accounting, without prompts or responses
5
+ Author-email: Luca <hello@luca.network>
6
+ License: Apache-2.0
7
+ Project-URL: Homepage, https://luca.network/developers
8
+ Project-URL: Documentation, https://luca.network/developers#litellm
9
+ Project-URL: Repository, https://github.com/luca-network/luca-litellm
10
+ Keywords: litellm,ai spend,accounting,finops,cost of revenue,llm cost,gateway
11
+ Classifier: License :: OSI Approved :: Apache Software License
12
+ Classifier: Programming Language :: Python :: 3
13
+ Classifier: Topic :: Office/Business :: Financial :: Accounting
14
+ Requires-Python: >=3.9
15
+ Description-Content-Type: text/markdown
16
+ License-File: LICENSE
17
+ License-File: NOTICE
18
+ Requires-Dist: litellm>=1.70
19
+ Requires-Dist: httpx>=0.27
20
+ Dynamic: license-file
21
+
22
+ # luca-litellm
23
+
24
+ Send the cost of every call through your LiteLLM proxy to Luca, so your finance team gets AI spend by customer, team and model, accrued and matched to each provider's invoice, with journal entries for the ERP.
25
+
26
+ Only the fields accounting needs leave your network: when the call was made, the provider and model, the cost and token counts, and the team, key, project, customer, tags and MCP server LiteLLM records with it. The plugin drops the prompt, the response, tool arguments and tool results before anything is sent. Email addresses are left out unless you turn them on.
27
+
28
+ ## Set up
29
+
30
+ 1. In Luca, add a **LiteLLM AI gateway** connection. Luca shows its webhook URL. Make an ingest token for it:
31
+ `openssl rand -hex 32`
32
+ 2. On the machine that runs the proxy:
33
+ ```
34
+ pip install luca-litellm
35
+ ```
36
+ 3. Add the plugin to the proxy's `config.yaml`:
37
+ ```yaml
38
+ litellm_settings:
39
+ callbacks: ["luca_litellm.handler"]
40
+ ```
41
+ 4. Set two environment variables for the proxy:
42
+ ```
43
+ LUCA_LITELLM_URL=<the webhook URL Luca shows>
44
+ LUCA_LITELLM_TOKEN=<the ingest token>
45
+ ```
46
+ 5. Restart the proxy.
47
+
48
+ The plugin sends calls in batches (every 10 seconds, or every 200 calls), outside the request path, so it adds no time to your calls. If Luca can't be reached, it keeps the calls and tries again.
49
+
50
+ ## Which customer a call served
51
+
52
+ LiteLLM records a customer for each call when your app passes `user` in the request, or when the call uses a key or team set up for that customer. Luca uses it to show AI cost by customer.
53
+
54
+ ## Options
55
+
56
+ | Variable | What it does |
57
+ |---|---|
58
+ | `LUCA_LITELLM_URL` | The connection's webhook URL (required) |
59
+ | `LUCA_LITELLM_TOKEN` | The connection's ingest token (required) |
60
+ | `LUCA_LITELLM_INCLUDE_EMAIL` | `true` to send the user's email address with each call (off by default) |
61
+
62
+ To use it from the LiteLLM Python SDK instead of the proxy:
63
+
64
+ ```python
65
+ import litellm
66
+ from luca_litellm import LucaLogger
67
+ litellm.callbacks = [LucaLogger(url="https://…", token="…")]
68
+ ```
69
+
70
+ ## Without installing a package
71
+
72
+ LiteLLM's built-in webhook callback can send the same calls. Luca shows that configuration on the connection too. With it, turn message logging off (`turn_off_message_logging: true`) so prompts and responses are not sent.
73
+
74
+ Apache-2.0. Questions: hello@luca.network
@@ -0,0 +1,11 @@
1
+ LICENSE
2
+ NOTICE
3
+ README.md
4
+ pyproject.toml
5
+ luca_litellm/__init__.py
6
+ luca_litellm.egg-info/PKG-INFO
7
+ luca_litellm.egg-info/SOURCES.txt
8
+ luca_litellm.egg-info/dependency_links.txt
9
+ luca_litellm.egg-info/requires.txt
10
+ luca_litellm.egg-info/top_level.txt
11
+ tests/test_luca_litellm.py
@@ -0,0 +1,2 @@
1
+ litellm>=1.70
2
+ httpx>=0.27
@@ -0,0 +1 @@
1
+ luca_litellm
@@ -0,0 +1,28 @@
1
+ [build-system]
2
+ requires = ["setuptools>=68"]
3
+ build-backend = "setuptools.build_meta"
4
+
5
+ [project]
6
+ name = "luca-litellm"
7
+ version = "0.1.0"
8
+ description = "Send the cost of each LiteLLM call to Luca for accounting, without prompts or responses"
9
+ readme = "README.md"
10
+ license = { text = "Apache-2.0" }
11
+ authors = [{ name = "Luca", email = "hello@luca.network" }]
12
+ requires-python = ">=3.9"
13
+ dependencies = ["litellm>=1.70", "httpx>=0.27"]
14
+ keywords = ["litellm", "ai spend", "accounting", "finops", "cost of revenue", "llm cost", "gateway"]
15
+ classifiers = [
16
+ "License :: OSI Approved :: Apache Software License",
17
+ "Programming Language :: Python :: 3",
18
+ "Topic :: Office/Business :: Financial :: Accounting",
19
+ ]
20
+
21
+ [project.urls]
22
+ Homepage = "https://luca.network/developers"
23
+ Documentation = "https://luca.network/developers#litellm"
24
+ Repository = "https://github.com/luca-network/luca-litellm"
25
+
26
+ [tool.setuptools]
27
+ packages = ["luca_litellm"]
28
+ license-files = ["LICENSE", "NOTICE"]
@@ -0,0 +1,4 @@
1
+ [egg_info]
2
+ tag_build =
3
+ tag_date = 0
4
+
@@ -0,0 +1,140 @@
1
+ """luca-litellm: what leaves the proxy, batching, retries, and a real LiteLLM call through the callback."""
2
+
3
+ import asyncio
4
+ import json
5
+ import os
6
+ import sys
7
+ import threading
8
+ from http.server import BaseHTTPRequestHandler, HTTPServer
9
+
10
+ import litellm
11
+ import pytest
12
+
13
+ from luca_litellm import LucaLogger, to_record
14
+
15
+ PAYLOAD = {
16
+ "id": "call-1", "status": "success", "call_type": "acompletion", "startTime": 1791072000.5, "endTime": 1791072001.0,
17
+ "model": "gpt-4o", "model_group": "gpt-4o", "custom_llm_provider": "openai",
18
+ "api_base": "https://myco.openai.azure.com/openai/deployments/prod?api-key=SECRETKEY",
19
+ "response_cost": 0.0123, "prompt_tokens": 120, "completion_tokens": 30, "total_tokens": 150, "cache_hit": False,
20
+ "end_user": "acme-prod", "request_tags": ["batch"],
21
+ "messages": [{"role": "user", "content": "secret prompt"}], "response": {"text": "secret answer"},
22
+ "model_parameters": {"temperature": 0}, "requester_ip_address": "10.0.0.8",
23
+ "metadata": {
24
+ "user_api_key_hash": "hash1", "user_api_key_team_id": "t_1", "user_api_key_team_alias": "Support",
25
+ "user_api_key_user_id": "u_9", "user_api_key_user_email": "person@myco.example",
26
+ "usage_object": {"prompt_tokens_details": {"cached_tokens": 20}, "completion_tokens": 30},
27
+ "mcp_tool_call_metadata": {"name": "search", "mcp_server_name": "github", "arguments": {"q": "secret"}, "result": {"r": "secret"}},
28
+ "requester_metadata": {"note": "secret"},
29
+ },
30
+ }
31
+
32
+
33
+ def test_only_the_fields_accounting_needs_leave_the_proxy():
34
+ r = to_record(PAYLOAD)
35
+ text = json.dumps(r)
36
+ assert "secret" not in text.lower() and "SECRETKEY" not in text, "no prompt, response, tool arguments or key"
37
+ assert "person@myco.example" not in text, "no email address unless asked for"
38
+ assert "10.0.0.8" not in text
39
+ assert r["api_base"] == "https://myco.openai.azure.com"
40
+ assert r["response_cost"] == 0.0123 and r["prompt_tokens"] == 120
41
+ assert r["end_user"] == "acme-prod" and r["request_tags"] == ["batch"]
42
+ assert r["metadata"]["user_api_key_team_alias"] == "Support"
43
+ assert r["metadata"]["usage_object"] == {"prompt_tokens_details": {"cached_tokens": 20}}
44
+ assert r["metadata"]["mcp_tool_call_metadata"] == {"mcp_server_name": "github", "name": "search"}
45
+ assert to_record(PAYLOAD, include_email=True)["metadata"]["user_api_key_user_email"] == "person@myco.example"
46
+ assert to_record({**PAYLOAD, "status": "failure"}) is None
47
+ assert to_record(None) is None and to_record({"status": "success"}) is None
48
+
49
+
50
+ class Receiver:
51
+ """A local stand-in for Luca's webhook: records each delivery, and can fail the first few."""
52
+
53
+ def __init__(self, fail_first=0, status=503):
54
+ self.deliveries, self.headers, self.fail_left, self.status = [], [], fail_first, status
55
+ outer = self
56
+
57
+ class H(BaseHTTPRequestHandler):
58
+ def do_POST(self):
59
+ body = self.rfile.read(int(self.headers["Content-Length"]))
60
+ if outer.fail_left > 0:
61
+ outer.fail_left -= 1
62
+ self.send_response(outer.status); self.end_headers(); return
63
+ outer.deliveries.append(json.loads(body)); outer.headers.append(dict(self.headers))
64
+ self.send_response(200); self.send_header("content-type", "application/json"); self.end_headers()
65
+ self.wfile.write(b'{"stored":1}')
66
+
67
+ def log_message(self, *a):
68
+ pass
69
+
70
+ self.server = HTTPServer(("127.0.0.1", 0), H)
71
+ self.url = f"http://127.0.0.1:{self.server.server_port}/v1/hooks/litellm/conn_1"
72
+ threading.Thread(target=self.server.serve_forever, daemon=True).start()
73
+
74
+ def close(self):
75
+ self.server.shutdown()
76
+
77
+
78
+ def test_batches_retries_and_never_sends_without_settings(monkeypatch):
79
+ monkeypatch.delenv("LUCA_LITELLM_URL", raising=False)
80
+ monkeypatch.delenv("LUCA_LITELLM_TOKEN", raising=False)
81
+ idle = LucaLogger()
82
+ idle.log_success_event({"standard_logging_object": PAYLOAD}, None, None, None)
83
+ idle.flush() # no URL or token: nothing is sent, nothing raises
84
+ assert len(idle._queue) == 1
85
+
86
+ rx = Receiver(fail_first=1)
87
+ try:
88
+ lg = LucaLogger(url=rx.url, token="t" * 40, batch_size=2)
89
+ lg.log_success_event({"standard_logging_object": PAYLOAD}, None, None, None)
90
+ lg.log_success_event({"standard_logging_object": PAYLOAD}, None, None, None) # the same call twice: kept once
91
+ assert rx.deliveries == [] and len(lg._queue) == 1
92
+ lg.log_success_event({"standard_logging_object": {**PAYLOAD, "id": "call-2"}}, None, None, None)
93
+ assert rx.deliveries == [], "Luca was down: the batch is kept"
94
+ assert len(lg._queue) == 2
95
+ lg.flush()
96
+ assert [r["id"] for r in rx.deliveries[0]] == ["call-1", "call-2"]
97
+ assert rx.headers[0]["Authorization"] == "Bearer " + "t" * 40
98
+ assert lg._queue == []
99
+ finally:
100
+ rx.close()
101
+
102
+ bad = Receiver(fail_first=1, status=401)
103
+ try:
104
+ lg = LucaLogger(url=bad.url, token="t" * 40)
105
+ lg.log_success_event({"standard_logging_object": PAYLOAD}, None, None, None)
106
+ lg.flush()
107
+ assert lg._queue == [], "a refused delivery (wrong token) is not retried for ever"
108
+ finally:
109
+ bad.close()
110
+
111
+
112
+ def test_a_real_litellm_call_through_the_callback():
113
+ rx = Receiver()
114
+ try:
115
+ lg = LucaLogger(url=rx.url, token="t" * 40, flush_seconds=1)
116
+
117
+ async def go():
118
+ await litellm.acompletion(
119
+ model="gpt-4o", messages=[{"role": "user", "content": "a secret question"}],
120
+ mock_response="a secret answer", callbacks=[lg], user="acme-prod",
121
+ metadata={"tags": ["batch"]},
122
+ )
123
+ for _ in range(50):
124
+ await asyncio.sleep(0.1)
125
+ if lg._queue or rx.deliveries:
126
+ break
127
+ await lg.async_flush()
128
+
129
+ asyncio.run(go())
130
+ sent = [r for d in rx.deliveries for r in d]
131
+ assert len(sent) == 1, sent
132
+ r = sent[0]
133
+ assert r["model"] == "gpt-4o" and r["status"] == "success" and r["prompt_tokens"] > 0
134
+ assert "secret" not in json.dumps(r)
135
+ out = os.environ.get("LUCA_LITELLM_SAMPLE")
136
+ if out: # the delivery, kept for luca-core's own test of reading it
137
+ with open(out, "w") as f:
138
+ json.dump(rx.deliveries[0], f, indent=1)
139
+ finally:
140
+ rx.close()