egma 0.2.0__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.
- egma-0.2.0/LICENSE +201 -0
- egma-0.2.0/PKG-INFO +398 -0
- egma-0.2.0/README.md +377 -0
- egma-0.2.0/pyproject.toml +68 -0
- egma-0.2.0/src/egma/__init__.py +34 -0
- egma-0.2.0/src/egma/mockable.py +1113 -0
- egma-0.2.0/src/egma/monitoring.py +320 -0
- egma-0.2.0/src/egma/seam.py +403 -0
egma-0.2.0/LICENSE
ADDED
|
@@ -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 Egma
|
|
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.
|
egma-0.2.0/PKG-INFO
ADDED
|
@@ -0,0 +1,398 @@
|
|
|
1
|
+
Metadata-Version: 2.4
|
|
2
|
+
Name: egma
|
|
3
|
+
Version: 0.2.0
|
|
4
|
+
Summary: Test and monitor LiveKit agents with Egma.
|
|
5
|
+
Keywords: livekit,voice,agents,testing,mock tools,egma
|
|
6
|
+
License-Expression: Apache-2.0
|
|
7
|
+
License-File: LICENSE
|
|
8
|
+
Classifier: Development Status :: 3 - Alpha
|
|
9
|
+
Classifier: Intended Audience :: Developers
|
|
10
|
+
Classifier: Programming Language :: Python :: 3
|
|
11
|
+
Classifier: Topic :: Software Development :: Testing
|
|
12
|
+
Requires-Dist: livekit-agents>=1.6.7,<1.7
|
|
13
|
+
Requires-Dist: openai>=2,<3
|
|
14
|
+
Requires-Dist: opentelemetry-exporter-otlp-proto-http>=1.39,<2
|
|
15
|
+
Requires-Dist: opentelemetry-sdk>=1.39,<2
|
|
16
|
+
Requires-Python: >=3.11
|
|
17
|
+
Project-URL: Homepage, https://github.com/egma-ai/egma
|
|
18
|
+
Project-URL: Repository, https://github.com/egma-ai/egma
|
|
19
|
+
Project-URL: Documentation, https://github.com/egma-ai/egma/tree/main/sdks/python#readme
|
|
20
|
+
Description-Content-Type: text/markdown
|
|
21
|
+
|
|
22
|
+
# Egma Python SDK
|
|
23
|
+
|
|
24
|
+
Test a LiveKit agent with mock tools and send its production spans to Egma.
|
|
25
|
+
|
|
26
|
+
The two functions are separate:
|
|
27
|
+
|
|
28
|
+
- `mockable(...)` lets Egma answer tools in an Egma simulation. It does
|
|
29
|
+
nothing in production.
|
|
30
|
+
- `monitor_livekit(...)` sends production LiveKit spans to the
|
|
31
|
+
Monitoring page. It does nothing in an Egma simulation.
|
|
32
|
+
|
|
33
|
+
Calling one function never enables or changes the other.
|
|
34
|
+
|
|
35
|
+
## Install
|
|
36
|
+
|
|
37
|
+
```bash
|
|
38
|
+
pip install 'egma>=0.2.0'
|
|
39
|
+
```
|
|
40
|
+
|
|
41
|
+
`0.2.0` is the floor because it is the first release that knows a
|
|
42
|
+
simulation by the room's name, and so the first that holds on every one of
|
|
43
|
+
the three dispatch paths. An unpinned install can resolve to a release
|
|
44
|
+
that looks in the dispatch metadata instead, where Egma no longer writes
|
|
45
|
+
anything — inert inside a real simulation on all three.
|
|
46
|
+
|
|
47
|
+
For a LiveKit agent on Python 3.11 or newer.
|
|
48
|
+
|
|
49
|
+
## Production monitoring
|
|
50
|
+
|
|
51
|
+
Open **Monitoring → Start monitoring** and choose **LiveKit**. That page
|
|
52
|
+
shows the same SDK, key, and helper steps below. It is instructions only: a
|
|
53
|
+
LiveKit agent pushes its own spans, so there is nothing to switch on in Egma.
|
|
54
|
+
|
|
55
|
+
Set the Egma API origin and an existing project API key in the agent's
|
|
56
|
+
environment:
|
|
57
|
+
|
|
58
|
+
```bash
|
|
59
|
+
export EGMA_URL=https://api.egma.ai
|
|
60
|
+
export EGMA_API_KEY=egma_sk_...
|
|
61
|
+
```
|
|
62
|
+
|
|
63
|
+
For self-hosted Egma, prefer the published API address, such as
|
|
64
|
+
`http://localhost:3100`. The web address also works because Egma forwards
|
|
65
|
+
`/v1/traces` to the API, but the API is one hop shorter. The agent process must
|
|
66
|
+
be able to make HTTP or HTTPS requests to the address you choose.
|
|
67
|
+
|
|
68
|
+
In a customer-hosted worker, set these values through your normal deployment
|
|
69
|
+
secret store. For LiveKit Cloud, place them in a gitignored secrets file and
|
|
70
|
+
apply it to the agent:
|
|
71
|
+
|
|
72
|
+
```bash
|
|
73
|
+
lk agent update-secrets --secrets-file=.env.monitoring
|
|
74
|
+
```
|
|
75
|
+
|
|
76
|
+
Use the same file with `lk agent create --secrets-file=.env.monitoring` for a
|
|
77
|
+
new deployment. LiveKit Cloud restarts the agent after a secret update.
|
|
78
|
+
|
|
79
|
+
Call `monitor_livekit` before `AgentSession.start`:
|
|
80
|
+
|
|
81
|
+
```python
|
|
82
|
+
from egma import monitor_livekit
|
|
83
|
+
from livekit import agents
|
|
84
|
+
from livekit.agents import Agent, AgentSession
|
|
85
|
+
|
|
86
|
+
|
|
87
|
+
async def entrypoint(ctx: agents.JobContext) -> None:
|
|
88
|
+
monitor_livekit(ctx)
|
|
89
|
+
await ctx.connect()
|
|
90
|
+
|
|
91
|
+
agent = Agent(instructions=INSTRUCTIONS, tools=[check_calendar])
|
|
92
|
+
session = AgentSession(stt=..., llm=..., tts=...)
|
|
93
|
+
await session.start(agent=agent, room=ctx.room)
|
|
94
|
+
```
|
|
95
|
+
|
|
96
|
+
You can pass the same values directly when environment variables are not the
|
|
97
|
+
right configuration source:
|
|
98
|
+
|
|
99
|
+
```python
|
|
100
|
+
monitor_livekit(ctx, endpoint="https://api.egma.ai", api_key=project_key)
|
|
101
|
+
```
|
|
102
|
+
|
|
103
|
+
Use the same call for an agent hosted in your cloud and an agent hosted in
|
|
104
|
+
LiveKit Cloud. The helper adds Egma to a compatible OpenTelemetry provider; it
|
|
105
|
+
does not remove LiveKit Cloud observability or another existing span
|
|
106
|
+
processor. It sends spans in batches and flushes the final batch when the
|
|
107
|
+
LiveKit job stops.
|
|
108
|
+
|
|
109
|
+
If this job runs in an Egma simulation room, the helper returns without
|
|
110
|
+
adding a production exporter and says so at `WARNING`. The simulation keeps
|
|
111
|
+
its own trace and does not appear a second time in Monitoring.
|
|
112
|
+
|
|
113
|
+
The helper reads the room's name for that and nothing else. Every Egma
|
|
114
|
+
simulation room is named `egma-sim-…`; a room named anything else gets the
|
|
115
|
+
production exporter. The name arrives with the job, so this decision costs
|
|
116
|
+
no network and happens before `ctx.connect()`.
|
|
117
|
+
|
|
118
|
+
After deployment, open **Monitoring**. Nothing needs to be confirmed in Egma —
|
|
119
|
+
the deployed helper is what sends spans, and the first production conversation
|
|
120
|
+
appears in the list as soon as it arrives.
|
|
121
|
+
|
|
122
|
+
The caller's phone, SIP, or WebRTC entry path does not change this setup. The
|
|
123
|
+
SDK does not guess that path from a LiveKit trace.
|
|
124
|
+
|
|
125
|
+
### Setup status
|
|
126
|
+
|
|
127
|
+
The helper stops immediately with one direct error when required settings are
|
|
128
|
+
missing or malformed, or when tracer providers conflict. It does not include
|
|
129
|
+
the key in that error.
|
|
130
|
+
|
|
131
|
+
After deployment, Monitoring stays empty until the first trace reaches Egma,
|
|
132
|
+
then lists each production conversation as it arrives. This release uses one-way
|
|
133
|
+
OTLP export. Egma cannot see a DNS, firewall, or network failure inside the
|
|
134
|
+
worker. Check the worker's OpenTelemetry logs and confirm that it can reach
|
|
135
|
+
`EGMA_URL` when nothing appears.
|
|
136
|
+
|
|
137
|
+
## Simulation mock tools
|
|
138
|
+
|
|
139
|
+
A simulation that reaches your real tools has real side effects: it books
|
|
140
|
+
the appointment, sends the message, charges the card. And a real backend
|
|
141
|
+
only ever shows you the branch its data happens to be on — "the calendar
|
|
142
|
+
is full", "the lookup fails", "the booking API errors" are where voice
|
|
143
|
+
agents die in production, and none of them can be ordered up from a real
|
|
144
|
+
backend on demand.
|
|
145
|
+
|
|
146
|
+
A **mock tool** answers for one of your agent's tools during a
|
|
147
|
+
simulation. It is authored in your Egma project, matched strictly by tool
|
|
148
|
+
name, and its answer may be a value, an error, or either one after a
|
|
149
|
+
declared delay so a mocked backend takes as long as the real one. This
|
|
150
|
+
package is the piece that lives in your own agent's process and lets Egma
|
|
151
|
+
answer.
|
|
152
|
+
|
|
153
|
+
Tools no mock tool covers run their real implementations, untouched.
|
|
154
|
+
Egma's record names which of your tools were covered and which were not,
|
|
155
|
+
so you always know whether a simulation was fully isolated.
|
|
156
|
+
|
|
157
|
+
### Use
|
|
158
|
+
|
|
159
|
+
One call, after the agent is built and before the session starts:
|
|
160
|
+
|
|
161
|
+
```python
|
|
162
|
+
from egma import mockable
|
|
163
|
+
```
|
|
164
|
+
|
|
165
|
+
```python
|
|
166
|
+
async def entrypoint(ctx: agents.JobContext) -> None:
|
|
167
|
+
agent = Agent(instructions=INSTRUCTIONS, tools=[check_calendar, book_appointment])
|
|
168
|
+
session = AgentSession(stt=..., llm=..., tts=...)
|
|
169
|
+
|
|
170
|
+
await mockable(agent, ctx, session)
|
|
171
|
+
|
|
172
|
+
await session.start(agent=agent, room=ctx.room)
|
|
173
|
+
```
|
|
174
|
+
|
|
175
|
+
That is the whole simulation integration.
|
|
176
|
+
|
|
177
|
+
### How it knows it is in a simulation
|
|
178
|
+
|
|
179
|
+
It reads the **room's name** off the job. Every room Egma conducts a
|
|
180
|
+
simulation in is named `egma-sim-…`, and that prefix is fixed. Nothing
|
|
181
|
+
else is read, nothing is asked, and no room is connected to find out.
|
|
182
|
+
|
|
183
|
+
The name is the signal because it is the one that arrives on every
|
|
184
|
+
dispatch path an agent can end up in an Egma room by — whether Egma
|
|
185
|
+
dispatches your worker by name, whether LiveKit walks an unnamed worker
|
|
186
|
+
into the room, and whether your own token endpoint puts the agent there.
|
|
187
|
+
A signal carried by an explicit dispatch arrives on only the first of
|
|
188
|
+
those.
|
|
189
|
+
|
|
190
|
+
In a simulation room, `mockable` connects the job with LiveKit's own
|
|
191
|
+
`JobContext.connect()` if your startup has not already done so, then finds
|
|
192
|
+
Egma among the room's participants: Egma joins as `egma-persona` or
|
|
193
|
+
`egma-persona-<simulation>`. On two of the three dispatch paths your
|
|
194
|
+
agent is in the room **before** Egma, so it waits for that participant,
|
|
195
|
+
for up to 45 seconds and without polling anything outside the room. That
|
|
196
|
+
is a long time to hold an agent before it greets anybody, and it is the
|
|
197
|
+
price of the wait being correct on every dispatch path rather than on one;
|
|
198
|
+
a simulation ordinarily pays a fraction of it. It does not reconnect an
|
|
199
|
+
already-connected room, and it never connects a production room.
|
|
200
|
+
|
|
201
|
+
If nobody by that name arrives, nothing is wrapped, your tools all run
|
|
202
|
+
their own implementations, and the reason is logged at `ERROR`. If two
|
|
203
|
+
participants answer to that name, the exchange is refused for the same
|
|
204
|
+
reason a room with two claimants has no knowable answer — and your tool
|
|
205
|
+
inventory is not sent to either of them.
|
|
206
|
+
|
|
207
|
+
Then `mockable` reports your agent's tools to Egma — names and schemas,
|
|
208
|
+
read off the agent object, so mock authoring starts from your real tool
|
|
209
|
+
names instead of your memory of them — and learns which of them this
|
|
210
|
+
simulation answers for. Calls to those tools go to Egma and come back with
|
|
211
|
+
the authored answer. Every one of them lands on the simulation's record
|
|
212
|
+
with its arguments, its answer, how long it took, and which mock tool
|
|
213
|
+
answered.
|
|
214
|
+
|
|
215
|
+
**In every other room it does nothing at all.** A room your own system
|
|
216
|
+
named — which is every production room — is a room where `mockable`
|
|
217
|
+
returns having touched nothing: no wrapper, no message, no connect. Your
|
|
218
|
+
tools are the same objects, called the same way, with no wrapper between
|
|
219
|
+
them and the model. Zero added latency, by construction rather than by
|
|
220
|
+
care. That property is a test in this package (`tests/test_inert.py`),
|
|
221
|
+
not a promise in this file.
|
|
222
|
+
|
|
223
|
+
Your job's **dispatch metadata is yours**. This SDK writes nothing into it
|
|
224
|
+
and reads nothing out of it — not one key, in any room, for any purpose.
|
|
225
|
+
Neither does Egma: on every dispatch path, both the room's metadata and
|
|
226
|
+
the dispatch's carry the string configured on the connection, byte for
|
|
227
|
+
byte. `json.loads(ctx.job.metadata)["your_key"]` reads the same thing in a
|
|
228
|
+
simulation that it reads in production.
|
|
229
|
+
|
|
230
|
+
### Where to call it
|
|
231
|
+
|
|
232
|
+
After the agent object exists and before `session.start`. The report of
|
|
233
|
+
your tools is the first thing said, so an Egma that is not in the room is
|
|
234
|
+
discovered before any tool call rather than half way through a test.
|
|
235
|
+
|
|
236
|
+
Keep one `mockable` call for the initial agent. The SDK follows LiveKit's
|
|
237
|
+
public handoff events and installs the same simulation couriers for each
|
|
238
|
+
selected `Agent` or `AgentTask` before that activity starts. Each handoff
|
|
239
|
+
also extends one cumulative tool census, so Egma's coverage record keeps
|
|
240
|
+
the tools found on earlier agents.
|
|
241
|
+
|
|
242
|
+
### What a call to a mocked tool does
|
|
243
|
+
|
|
244
|
+
- Goes to Egma over the same LiveKit room. No new endpoint, no new
|
|
245
|
+
credential, nothing new to expose. If needed, `mockable` connects that
|
|
246
|
+
room through the job context before sending the first RPC.
|
|
247
|
+
- Comes back with the authored answer, or raises the authored error as
|
|
248
|
+
the tool's own error, so your agent handles it exactly as it would
|
|
249
|
+
handle a real backend failing.
|
|
250
|
+
- **Falls open** if Egma turns out not to be reachable: your real tool
|
|
251
|
+
runs, and the agent behaves as it would with this package uninstalled.
|
|
252
|
+
- **Never waits forever.** Every branch ends in an answer, an error the
|
|
253
|
+
model can hear, or your own tool running.
|
|
254
|
+
|
|
255
|
+
A tool you attach to the agent *after* calling `mockable` is still
|
|
256
|
+
intercepted on its first call — Egma's answers are held by name. Its
|
|
257
|
+
arguments may be incomplete on the record, and Egma marks that call so
|
|
258
|
+
you can see it.
|
|
259
|
+
|
|
260
|
+
### Logging
|
|
261
|
+
|
|
262
|
+
Everything this package says goes to the `egma` logger. It is worth
|
|
263
|
+
having on at `INFO` the first time you wire an agent up: the line after
|
|
264
|
+
the census names how many tools you have and how many Egma answers for.
|
|
265
|
+
|
|
266
|
+
`ERROR` is reserved for a simulation that could not be isolated and can
|
|
267
|
+
be acted on — no Egma participant arrived in a simulation room, two
|
|
268
|
+
participants claimed to be Egma, or the two halves do not speak the same
|
|
269
|
+
version of the exchange. None of those lines is reachable in a production
|
|
270
|
+
room.
|
|
271
|
+
|
|
272
|
+
## Before you install anything: the interim recipe
|
|
273
|
+
|
|
274
|
+
You can get isolation today with no Egma code at all, using LiveKit's own
|
|
275
|
+
`mock_tools` and a guard you write yourself:
|
|
276
|
+
|
|
277
|
+
```python
|
|
278
|
+
from livekit.agents import mock_tools
|
|
279
|
+
|
|
280
|
+
|
|
281
|
+
def in_a_simulation(ctx: agents.JobContext) -> bool:
|
|
282
|
+
return ctx.job.room.name.startswith("egma-sim-")
|
|
283
|
+
|
|
284
|
+
|
|
285
|
+
async def entrypoint(ctx: agents.JobContext) -> None:
|
|
286
|
+
await ctx.connect()
|
|
287
|
+
agent = Agent(instructions=INSTRUCTIONS, tools=[check_calendar])
|
|
288
|
+
session = AgentSession(stt=..., llm=..., tts=...)
|
|
289
|
+
|
|
290
|
+
if in_a_simulation(ctx):
|
|
291
|
+
mock_tools(
|
|
292
|
+
type(agent),
|
|
293
|
+
{"check_calendar": lambda day: "No free slots on that day."},
|
|
294
|
+
session=session,
|
|
295
|
+
)
|
|
296
|
+
|
|
297
|
+
await session.start(agent=agent, room=ctx.room)
|
|
298
|
+
```
|
|
299
|
+
|
|
300
|
+
A room your own system named means the guard is false and nothing is
|
|
301
|
+
wrapped. Note where that safety comes from, because it is not where
|
|
302
|
+
`mockable`'s comes from: this guard fires on a prefix being *present*
|
|
303
|
+
rather than on Egma being *absent*. So it is true in any room whose name
|
|
304
|
+
begins `egma-sim-`, including one Egma is not in — `mockable` waits for
|
|
305
|
+
Egma's participant and gives up out loud, while this guard has nobody to
|
|
306
|
+
wait for and simply mocks. Refuse that prefix wherever your own side mints
|
|
307
|
+
production tokens, or a production room named to look like a simulation
|
|
308
|
+
runs your canned answers against a live caller.
|
|
309
|
+
|
|
310
|
+
**Write the guard on the room name, and not on `egmaIdentity` in
|
|
311
|
+
`ctx.job.metadata`.** That key is in no simulation room at all: your
|
|
312
|
+
dispatch metadata is yours, and Egma writes nothing into it on any
|
|
313
|
+
dispatch path. A guard on it does not raise — it quietly fails to fire,
|
|
314
|
+
and every real tool it was meant to hold back runs inside a simulation.
|
|
315
|
+
|
|
316
|
+
What it cannot do is the rest of the job. One canned world for every
|
|
317
|
+
test, so you cannot write "the calendar is full" as a *test* — you would
|
|
318
|
+
be editing your agent's source to change a test's data. Nothing about
|
|
319
|
+
those calls reaches Egma's record: no arguments, no answers, no timings,
|
|
320
|
+
no coverage stamp, so graders that read tool facts have nothing to read.
|
|
321
|
+
No declared delay, so latency numbers from a mocked run flatter you.
|
|
322
|
+
|
|
323
|
+
**And the honest caveat: that guard couples your agent's source to how
|
|
324
|
+
Egma announces itself.** The room-name prefix is a stated contract rather
|
|
325
|
+
than an implementation detail, so it is the safe thing to key off — but
|
|
326
|
+
the *rest* of the mechanism is not: where Egma sits in the room, what
|
|
327
|
+
it is called, how long it takes to arrive, and what a room with two
|
|
328
|
+
claimants means are all Egma's to evolve. That is precisely what
|
|
329
|
+
`mockable` exists to own, and your side stays one line.
|
|
330
|
+
|
|
331
|
+
Use the recipe as the bridge, not as the small tier.
|
|
332
|
+
|
|
333
|
+
## Compatibility
|
|
334
|
+
|
|
335
|
+
The `egma-sim-` room-name prefix is a stated contract, not an internal
|
|
336
|
+
detail. Every room Egma opens for a simulation begins with it, on every
|
|
337
|
+
dispatch path, and it is there to be relied on and to be allowlisted in
|
|
338
|
+
a token endpoint. This SDK keys its whole simulation/production decision
|
|
339
|
+
off it.
|
|
340
|
+
|
|
341
|
+
Every Egma deployment names its rooms that way, so this package works
|
|
342
|
+
against a self-hosted Egma on whatever schedule its owner upgrades it.
|
|
343
|
+
Nothing else is consulted — in particular, neither metadata channel,
|
|
344
|
+
which carries your own configured JSON and nothing of Egma's.
|
|
345
|
+
|
|
346
|
+
This package pins `livekit-agents` to one minor version
|
|
347
|
+
(`>=1.6.7,<1.7`), and that is deliberate. Interception uses LiveKit's
|
|
348
|
+
testing API. Monitoring also reads LiveKit's current dynamic tracer
|
|
349
|
+
provider because the public telemetry API has a setter but no getter.
|
|
350
|
+
The fixture and live tests verify both seams before the supported range
|
|
351
|
+
changes.
|
|
352
|
+
|
|
353
|
+
The package also keeps the OpenAI Python package on major version 2.
|
|
354
|
+
LiveKit Agents 1.6 does not support OpenAI 3.
|
|
355
|
+
|
|
356
|
+
`mock_tools` writes into a side table LiveKit keeps per session. It does
|
|
357
|
+
not touch your agent, your agent's class, or your tool registry, and the
|
|
358
|
+
model keeps seeing your real tool schemas throughout. Never calling it
|
|
359
|
+
leaves everything byte for byte as it was, which is what makes the inert
|
|
360
|
+
path literal.
|
|
361
|
+
|
|
362
|
+
If that API ever moves, the documented fallback is
|
|
363
|
+
`Agent.update_tools(...)` with `function_tool(raw_schema=...)` wrappers
|
|
364
|
+
built from the real tools' schemas. It is a heavier mechanism — it takes
|
|
365
|
+
responsibility for schema fidelity, for the real implementations, and for
|
|
366
|
+
re-wrapping after a handoff — which is exactly why it is the fallback and
|
|
367
|
+
not the mechanism.
|
|
368
|
+
|
|
369
|
+
## Developing this package
|
|
370
|
+
|
|
371
|
+
The toolchain is [uv](https://docs.astral.sh/uv/).
|
|
372
|
+
|
|
373
|
+
```bash
|
|
374
|
+
uv sync
|
|
375
|
+
uv run ruff check src tests
|
|
376
|
+
uv run pytest
|
|
377
|
+
```
|
|
378
|
+
|
|
379
|
+
The normal suite is hermetic: no LiveKit server, no project, and no external
|
|
380
|
+
network. One monitoring test uses a local HTTP collector. One live test skips
|
|
381
|
+
visibly when the environment is silent, naming what it needs. To run it, point
|
|
382
|
+
it at a LiveKit project:
|
|
383
|
+
|
|
384
|
+
```bash
|
|
385
|
+
TEST_LIVEKIT_URL=wss://... \
|
|
386
|
+
TEST_LIVEKIT_API_KEY=... TEST_LIVEKIT_API_SECRET=... \
|
|
387
|
+
TEST_MODEL_API_KEY=... \
|
|
388
|
+
uv run pytest tests/test_live_mockable.py -v
|
|
389
|
+
```
|
|
390
|
+
|
|
391
|
+
Each name falls back to the plain one the tool's own CLI reads —
|
|
392
|
+
`LIVEKIT_URL`, `LIVEKIT_API_KEY`, `LIVEKIT_API_SECRET`, `OPENAI_API_KEY`
|
|
393
|
+
— so one environment serves this and whatever else you run beside it.
|
|
394
|
+
`TEST_MODEL_NAME` picks the model; it defaults to `gpt-4o-mini`.
|
|
395
|
+
|
|
396
|
+
## License
|
|
397
|
+
|
|
398
|
+
Apache-2.0, with the rest of Egma.
|