langchain-plimsoll 0.28.0__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 Carroll Guertin
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 +1,223 @@
1
+ Metadata-Version: 2.4
2
+ Name: langchain-plimsoll
3
+ Version: 0.28.0
4
+ Summary: A LangChain tool and a Deep Agents sandbox backend that run agent-written code in plimsoll, a self-hosted sandbox that refuses to run below the isolation you require
5
+ License-Expression: Apache-2.0
6
+ Project-URL: Homepage, https://github.com/plimsollmark/plimsoll/tree/main/clients/python/langchain-plimsoll
7
+ Project-URL: Source, https://github.com/plimsollmark/plimsoll/tree/main/clients/python/langchain-plimsoll
8
+ Project-URL: Issues, https://github.com/plimsollmark/plimsoll/issues
9
+ Keywords: langchain,langgraph,deepagents,sandbox,code execution,agents,gvisor
10
+ Classifier: Programming Language :: Python :: 3
11
+ Classifier: Programming Language :: Python :: 3 :: Only
12
+ Classifier: Typing :: Typed
13
+ Requires-Python: >=3.11
14
+ Description-Content-Type: text/markdown
15
+ License-File: LICENSE
16
+ Requires-Dist: langchain-core==1.8.0
17
+ Requires-Dist: plimsoll-client==0.28.0
18
+ Provides-Extra: deepagents
19
+ Requires-Dist: deepagents==0.7.38; extra == "deepagents"
20
+ Dynamic: license-file
21
+
22
+ # langchain-plimsoll
23
+
24
+ An agent that writes code has to run it somewhere, and "somewhere" is usually your machine or a hosted sandbox you
25
+ cannot inspect. This package runs a LangChain, LangGraph or Deep Agents agent's code in plimsoll, a sandbox service you
26
+ run yourself, and it has no fallback: if the sandbox is weaker than you asked for, nothing runs.
27
+
28
+ It has two parts:
29
+
30
+ - **`PlimsollCodeTool`**, a LangChain tool (`plimsoll_run_code`) that runs each call in a fresh sandbox. LangGraph's
31
+ `ToolNode`, `create_agent` and any graph of your own call it like any other LangChain tool.
32
+ - **`PlimsollSandbox`**, a Deep Agents sandbox backend over one plimsoll session: one sandbox kept for the agent's whole
33
+ run, so the files it writes and the shell's directory are there for its next command. Deep Agents gives the agent its
34
+ `execute` tool and its file tools (`ls`, `read_file`, `write_file`, `edit_file`, `glob`, `grep`) on top of it
35
+ ([Deep Agents sandboxes (EXTERNAL · official docs ↗)](https://docs.langchain.com/oss/python/deepagents/sandboxes)).
36
+
37
+ ## Install
38
+
39
+ ```sh
40
+ pip install langchain-plimsoll # the tool: needs only langchain-core
41
+ pip install 'langchain-plimsoll[deepagents]' # and the Deep Agents backend
42
+ ```
43
+
44
+ The tool needs only `langchain-core`. The example below builds its agent with `create_agent`, which is in `langchain`:
45
+ `pip install langchain` for it.
46
+
47
+ It needs a running `plimsolld`, the plimsoll daemon, reachable from the agent
48
+ ([getting started (EXTERNAL · source repo ↗)](https://github.com/plimsollmark/plimsoll/blob/main/docs/getting-started.md)).
49
+ The tool's default language is Python, which on docker needs a project image with `python3` in it
50
+ (`SANDBOX_DOCKER_PROJECT_IMAGE=plimsoll/sandbox-python:latest`): on the default image every Python call comes back
51
+ refused. Python 3.11 or later, as langchain-core requires.
52
+
53
+ ## Use the tool
54
+
55
+ ```python
56
+ from langchain.agents import create_agent
57
+ from langchain_plimsoll import Client, CodeExecutor, PlimsollCodeTool
58
+
59
+ executor = CodeExecutor(Client("https://plimsoll.internal:8746", token=TOKEN))
60
+ agent = create_agent(model, tools=[PlimsollCodeTool(executor)])
61
+ ```
62
+
63
+ The agent passes the tool code (Python or JavaScript) and optional input files, and gets back JSON saying whether the code
64
+ ran (yes, no with the reason, or unknown), the output, the exit code, whether anything was cut short or timed out, the
65
+ isolation tier the code ran behind, and the SHA-256 of the run's record.
66
+
67
+ ## Use the Deep Agents backend
68
+
69
+ ```python
70
+ from deepagents import create_deep_agent
71
+ from langchain_plimsoll import Client, PlimsollSandbox
72
+
73
+ with PlimsollSandbox.open(Client("https://plimsoll.internal:8746", token=TOKEN)) as backend:
74
+ agent = create_deep_agent(model=model, backend=backend, system_prompt=f"Work in {backend.root}.")
75
+ agent.invoke({"messages": [{"role": "user", "content": "Write and test a CSV parser."}]})
76
+ ```
77
+
78
+ Deep Agents does not tell the model where to work; `backend.root` is the sandbox's work directory (`/work` on docker),
79
+ which is where the agent's files belong. Leaving the `with` block closes the session, and the sandbox and its files are
80
+ gone.
81
+
82
+ What the backend needs and does:
83
+
84
+ - **Python 3 in the sandbox.** Deep Agents' file tools are small `python3` scripts it runs through `execute`. `open`
85
+ refuses a daemon that does not state Python before any sandbox is made; on docker that is the `plimsoll/sandbox-python`
86
+ image (`SANDBOX_DOCKER_PROJECT_IMAGE`).
87
+ - **Searches that never answer "no matches" by mistake.** Deep Agents' `grep` sends GNU grep options with its errors
88
+ discarded. plimsoll's docker images are Alpine Linux, whose `grep` is BusyBox's and has neither option, so every such
89
+ search would come back empty. When the sandbox's `grep` is not GNU grep, the backend sends each search through Deep
90
+ Agents' own `python3` search instead, which reports its errors. It skips hidden files, as Python's glob does.
91
+ - **A refusal stops the run.** If plimsoll refuses a command (the daemon is at capacity, the session has ended, the tier
92
+ is too low), or a command may have run without an answer to show for it, the backend raises. It never hands the model a
93
+ refusal dressed as command output, because Deep Agents reads that output as data: its `ls` reads text it cannot parse
94
+ as an empty directory. A command that runs and fails is a normal result with its exit code.
95
+ - **A thread resumed on another worker keeps its sandbox.** Open the backend with a checkpoint store and a key, and
96
+ attach to it from the worker that resumes the thread:
97
+
98
+ ```python
99
+ from plimsoll_client import MemoryCheckpointStore # for tests: use storage only your app can write
100
+
101
+ key = f"{user_id}/{thread_id}" # from identities your app verified, never from model or user text
102
+ backend = PlimsollSandbox.open(Client(url, token, daemon_name="prod"), checkpoint_store=store, checkpoint_key=key)
103
+ # On the worker that resumes the thread:
104
+ backend = PlimsollSandbox.attach(Client(url, token, daemon_name="prod"), key, store=store)
105
+ ```
106
+
107
+ The second worker gets the same sandbox and its files, and the first is refused from then on. A store is an object
108
+ with `load` and an atomic `compare_and_set`; every call passes through it, so two workers never send at once, and a
109
+ worker that died mid-command leaves a record the next one takes over from (if that command ran with its answer lost,
110
+ the next one's first command may not repeat it). A missing key raises `LookupError`, never a new sandbox. The daemon
111
+ needs callers with tokens: open development mode refuses the takeover, and anyone holding the token and a session's run
112
+ records can take it over. The rules:
113
+ [carrying a session on from another worker (EXTERNAL · source repo ↗)](https://github.com/plimsollmark/plimsoll/blob/main/docs/sessions.md#carrying-a-session-on-from-another-worker).
114
+ When the session ends (the daemon's lifetime or idle limit), every later call raises; the backend never opens a fresh
115
+ sandbox in its place, since the agent would take it for the one holding its files.
116
+ - **Every command has a budget.** A command with no timeout gets the backend's default (120 s); `timeout=0` gets the
117
+ daemon's longest; a command past its budget answers exit 124 with what it printed. A process left in the background is
118
+ stopped when its command ends.
119
+
120
+ ## Retries
121
+
122
+ LangGraph and LangChain retry nothing unless you ask: retries are off by default in both. You turn them on with a
123
+ `RetryPolicy` on a graph node (or on every node, with `set_node_defaults`), or with `ToolRetryMiddleware` on an agent.
124
+ Once on, both frameworks' default rules retry a plimsoll error, including one after which the code may already have run,
125
+ so a retry can run it a second time: a file appended to twice, a request sent twice. `retry_on` is the rule to give them
126
+ instead. It retries only a refusal plimsoll marked as having run nothing, and only when another try may pass: the daemon
127
+ was at capacity, or could not be reached before anything was sent. Every other error is left to stop the run.
128
+
129
+ ```python
130
+ from langchain.agents.middleware import ToolRetryMiddleware
131
+ from langgraph.types import RetryPolicy
132
+ from langchain_plimsoll import retry_on
133
+
134
+ # A graph node that calls plimsoll
135
+ builder.add_node("run", node, retry_policy=RetryPolicy(retry_on=retry_on))
136
+
137
+ # A Deep Agents agent: its execute and file tools
138
+ agent = create_deep_agent(model=model, backend=backend, middleware=[ToolRetryMiddleware(retry_on=retry_on)])
139
+ ```
140
+
141
+ - **Where it matters.** The Deep Agents backend raises on a refusal (above), and so do `Client` and `CodeExecutor`
142
+ called from a node of your own: that is where a retry rule meets plimsoll. `PlimsollCodeTool` never raises for a
143
+ refusal or a lost answer; it answers with them as its JSON (`ran` false or `"unknown"`), so no retry rule sees them.
144
+ - **Which refusals.** Every plimsoll error says whether anything ran: a refusal raised before any code was handed to the
145
+ sandbox carries a reason, and an error without one may have followed execution
146
+ ([what comes back, and what it means (EXTERNAL · source repo ↗)](https://github.com/plimsollmark/plimsoll/blob/main/docs/run-results.md)).
147
+ `retry_on` says yes to the reasons `capacity` (the daemon was full or over a rate limit and turned the run away) and
148
+ `environment` (the daemon or its token check could not be reached before anything was sent, or a session's sandbox
149
+ could not be read back before the call). It says no to the others, since the same daemon answers them the same way:
150
+ `request`, `permission`, `protocol`, `unsupported` and `isolation`. One `environment` cause, an image outside your
151
+ software rule, fails the same way each time; a retry of it costs the policy's few attempts and runs nothing.
152
+ - **Only plimsoll's errors.** It says no to every exception that is not plimsoll's (every plimsoll error is a
153
+ `plimsoll_client.PlimsollError`). For a node that also calls a service whose errors you want retried, decide both in
154
+ one rule: `RetryPolicy(retry_on=lambda e: retry_on(e) if isinstance(e, PlimsollError) else your_rule(e))`. A second
155
+ `RetryPolicy` on the same node does not do this: LangGraph uses the first policy whose rule says yes, so a later policy
156
+ with LangGraph's default rule would still retry plimsoll's errors.
157
+
158
+ ## What it does, and what it refuses to do
159
+
160
+ - **No fallback.** Both parts require the `kernel` isolation tier by default (gVisor, a user-space kernel, or a virtual
161
+ machine). Against a daemon that states less, they refuse before any code is sent. Lowering the floor is a choice you
162
+ make in code: `CodeExecutor(client, minimum_isolation="container")`, `PlimsollSandbox.open(client,
163
+ minimum_isolation="container")`.
164
+ - **A checked record of every run.** Each answer carries a run record, the daemon's statement of what was sent, what came
165
+ back and where it ran, and the client recomputes its digests before handing the result on.
166
+ - **One web API, if you grant one, for the tool.** `CodeExecutor(client, grant_profile="lights", api_description=...)`,
167
+ with a profile the operator configured on the daemon and the description `plimsoll-specgen` wrote for it, lets the
168
+ tool's code call that profile's routes, and nothing else, through plimsoll's broker, which adds the API's key outside
169
+ the sandbox. JavaScript works on every provider that gives a project run a grant; Python too on docker and openshell,
170
+ through `import plimsoll_host`. The tool reads the daemon's answer when it is built, offers only the languages that
171
+ can call the API there, and tells the model the routes; the model never names a profile. The code can do whatever an
172
+ allowed route permits and read whatever it returns
173
+ ([capability grants (EXTERNAL · source repo ↗)](https://github.com/plimsollmark/plimsoll/blob/main/docs/capability-grants.md#from-an-agent-kits-code-tool)).
174
+ The Deep Agents backend's calls take no grant.
175
+ - **Tested against LangChain's own suites.** The tool passes LangChain's standard tool tests (`langchain-tests`). The
176
+ backend passes LangChain's standard sandbox tests on a real plimsoll session, except seven cases where
177
+ langchain-tests 1.1.9 and deepagents 0.7.38 disagree with each other (six about what `glob` returns, one about writing
178
+ over an existing file); they fail the same way for any backend built on Deep Agents' `BaseSandbox`.
179
+
180
+ ## Tracing
181
+
182
+ If you turn on LangSmith tracing, LangChain sends each tool call's inputs and outputs, which here are the model's code
183
+ and what it printed, to LangSmith. That is LangChain's switch, not this package's; LangSmith's `hide_inputs` and
184
+ `hide_outputs` settings turn it off
185
+ ([LangSmith tracing (EXTERNAL · official docs ↗)](https://docs.langchain.com/langsmith/mask-inputs-outputs)).
186
+
187
+ `PlimsollCodeTool` adds two things of its own, and neither carries the code or its output:
188
+
189
+ - **One ID in both logs.** Each call sends the LangChain run ID of the tool call (the ID LangSmith shows for that run) as
190
+ the run's `trace_id`, which the daemon writes on the run's audit line and does nothing else with. To find what the
191
+ daemon did for a traced tool call, search its log for the run's ID. The daemon accepts 1 to 64 characters of letters,
192
+ digits and `.`, `_`, `:` and `-`; LangChain's IDs are UUIDs, which fit. An ID you set yourself (`run_id` in the
193
+ config) that does not fit is not sent at all, never cut short, since a shortened ID would match nothing.
194
+ - **Labels on the tool call's run.** When the call is traced, the tool adds metadata to its run: `plimsoll_record_sha256`
195
+ (the run record's digest, which the client checked), `plimsoll_isolation` (the tier the code ran behind) and
196
+ `plimsoll_isolation_qualification` (the sentence a provider states beside its tier, when it states one),
197
+ `plimsoll_dispatched` (`yes`, `no` or `unknown`: whether the run reached the provider, or an error left that open),
198
+ `plimsoll_refused` (the reason, when it did not) and `plimsoll_outcome` (how the run ended: `completed`,
199
+ `setup_failed`, `timed_out`, `protocol_error` or `runner_failed`). Each value is checked against the values plimsoll
200
+ can state and left out if it does not match, so nothing the code printed can reach these fields. In LangSmith you can
201
+ then filter for runs that were refused, ran below a tier, or may have run without an answer. With tracing off the tool
202
+ adds nothing, and it imports no part of LangSmith that is not already loaded.
203
+
204
+ ## What it is not
205
+
206
+ - **Not a hosted service.** You run `plimsolld` on your own machine or cluster, with a provider you pick: docker under
207
+ gVisor, an E2B or Docker Cloud microVM, a Kubernetes agent-sandbox pod, and others
208
+ ([isolation tiers (EXTERNAL · source repo ↗)](https://github.com/plimsollmark/plimsoll/blob/main/docs/isolation-tiers.md)).
209
+ The Deep Agents backend needs a provider with sessions: docker, E2B or OpenShell.
210
+ - **Not attestation.** A reported tier is the daemon's configuration and provider evidence plus a startup check, not
211
+ hardware proof.
212
+
213
+ ## Versions
214
+
215
+ This package's version moves with plimsoll's releases, and each version requires the `plimsoll-client` release of the same
216
+ number, which holds the code (`plimsoll_client.langchain` and `plimsoll_client.deepagents`). `Client`, `CodeExecutor`,
217
+ `PlimsollCodeTool` and `PlimsollSandbox` here are plimsoll-client's own classes, gathered so an agent imports from one
218
+ place. It is tested with, and installs with, exactly langchain-core 1.8.0, and its `deepagents` extra with exactly
219
+ deepagents 0.7.38: pip installs those versions beside it and refuses others, and each new release of either is tested
220
+ before its pin moves.
221
+
222
+ Licensed under the Apache License 2.0. Source:
223
+ [plimsoll on GitHub (EXTERNAL · source repo ↗)](https://github.com/plimsollmark/plimsoll/tree/main/clients/python/langchain-plimsoll).
@@ -0,0 +1,202 @@
1
+ # langchain-plimsoll
2
+
3
+ An agent that writes code has to run it somewhere, and "somewhere" is usually your machine or a hosted sandbox you
4
+ cannot inspect. This package runs a LangChain, LangGraph or Deep Agents agent's code in plimsoll, a sandbox service you
5
+ run yourself, and it has no fallback: if the sandbox is weaker than you asked for, nothing runs.
6
+
7
+ It has two parts:
8
+
9
+ - **`PlimsollCodeTool`**, a LangChain tool (`plimsoll_run_code`) that runs each call in a fresh sandbox. LangGraph's
10
+ `ToolNode`, `create_agent` and any graph of your own call it like any other LangChain tool.
11
+ - **`PlimsollSandbox`**, a Deep Agents sandbox backend over one plimsoll session: one sandbox kept for the agent's whole
12
+ run, so the files it writes and the shell's directory are there for its next command. Deep Agents gives the agent its
13
+ `execute` tool and its file tools (`ls`, `read_file`, `write_file`, `edit_file`, `glob`, `grep`) on top of it
14
+ ([Deep Agents sandboxes (EXTERNAL · official docs ↗)](https://docs.langchain.com/oss/python/deepagents/sandboxes)).
15
+
16
+ ## Install
17
+
18
+ ```sh
19
+ pip install langchain-plimsoll # the tool: needs only langchain-core
20
+ pip install 'langchain-plimsoll[deepagents]' # and the Deep Agents backend
21
+ ```
22
+
23
+ The tool needs only `langchain-core`. The example below builds its agent with `create_agent`, which is in `langchain`:
24
+ `pip install langchain` for it.
25
+
26
+ It needs a running `plimsolld`, the plimsoll daemon, reachable from the agent
27
+ ([getting started (EXTERNAL · source repo ↗)](https://github.com/plimsollmark/plimsoll/blob/main/docs/getting-started.md)).
28
+ The tool's default language is Python, which on docker needs a project image with `python3` in it
29
+ (`SANDBOX_DOCKER_PROJECT_IMAGE=plimsoll/sandbox-python:latest`): on the default image every Python call comes back
30
+ refused. Python 3.11 or later, as langchain-core requires.
31
+
32
+ ## Use the tool
33
+
34
+ ```python
35
+ from langchain.agents import create_agent
36
+ from langchain_plimsoll import Client, CodeExecutor, PlimsollCodeTool
37
+
38
+ executor = CodeExecutor(Client("https://plimsoll.internal:8746", token=TOKEN))
39
+ agent = create_agent(model, tools=[PlimsollCodeTool(executor)])
40
+ ```
41
+
42
+ The agent passes the tool code (Python or JavaScript) and optional input files, and gets back JSON saying whether the code
43
+ ran (yes, no with the reason, or unknown), the output, the exit code, whether anything was cut short or timed out, the
44
+ isolation tier the code ran behind, and the SHA-256 of the run's record.
45
+
46
+ ## Use the Deep Agents backend
47
+
48
+ ```python
49
+ from deepagents import create_deep_agent
50
+ from langchain_plimsoll import Client, PlimsollSandbox
51
+
52
+ with PlimsollSandbox.open(Client("https://plimsoll.internal:8746", token=TOKEN)) as backend:
53
+ agent = create_deep_agent(model=model, backend=backend, system_prompt=f"Work in {backend.root}.")
54
+ agent.invoke({"messages": [{"role": "user", "content": "Write and test a CSV parser."}]})
55
+ ```
56
+
57
+ Deep Agents does not tell the model where to work; `backend.root` is the sandbox's work directory (`/work` on docker),
58
+ which is where the agent's files belong. Leaving the `with` block closes the session, and the sandbox and its files are
59
+ gone.
60
+
61
+ What the backend needs and does:
62
+
63
+ - **Python 3 in the sandbox.** Deep Agents' file tools are small `python3` scripts it runs through `execute`. `open`
64
+ refuses a daemon that does not state Python before any sandbox is made; on docker that is the `plimsoll/sandbox-python`
65
+ image (`SANDBOX_DOCKER_PROJECT_IMAGE`).
66
+ - **Searches that never answer "no matches" by mistake.** Deep Agents' `grep` sends GNU grep options with its errors
67
+ discarded. plimsoll's docker images are Alpine Linux, whose `grep` is BusyBox's and has neither option, so every such
68
+ search would come back empty. When the sandbox's `grep` is not GNU grep, the backend sends each search through Deep
69
+ Agents' own `python3` search instead, which reports its errors. It skips hidden files, as Python's glob does.
70
+ - **A refusal stops the run.** If plimsoll refuses a command (the daemon is at capacity, the session has ended, the tier
71
+ is too low), or a command may have run without an answer to show for it, the backend raises. It never hands the model a
72
+ refusal dressed as command output, because Deep Agents reads that output as data: its `ls` reads text it cannot parse
73
+ as an empty directory. A command that runs and fails is a normal result with its exit code.
74
+ - **A thread resumed on another worker keeps its sandbox.** Open the backend with a checkpoint store and a key, and
75
+ attach to it from the worker that resumes the thread:
76
+
77
+ ```python
78
+ from plimsoll_client import MemoryCheckpointStore # for tests: use storage only your app can write
79
+
80
+ key = f"{user_id}/{thread_id}" # from identities your app verified, never from model or user text
81
+ backend = PlimsollSandbox.open(Client(url, token, daemon_name="prod"), checkpoint_store=store, checkpoint_key=key)
82
+ # On the worker that resumes the thread:
83
+ backend = PlimsollSandbox.attach(Client(url, token, daemon_name="prod"), key, store=store)
84
+ ```
85
+
86
+ The second worker gets the same sandbox and its files, and the first is refused from then on. A store is an object
87
+ with `load` and an atomic `compare_and_set`; every call passes through it, so two workers never send at once, and a
88
+ worker that died mid-command leaves a record the next one takes over from (if that command ran with its answer lost,
89
+ the next one's first command may not repeat it). A missing key raises `LookupError`, never a new sandbox. The daemon
90
+ needs callers with tokens: open development mode refuses the takeover, and anyone holding the token and a session's run
91
+ records can take it over. The rules:
92
+ [carrying a session on from another worker (EXTERNAL · source repo ↗)](https://github.com/plimsollmark/plimsoll/blob/main/docs/sessions.md#carrying-a-session-on-from-another-worker).
93
+ When the session ends (the daemon's lifetime or idle limit), every later call raises; the backend never opens a fresh
94
+ sandbox in its place, since the agent would take it for the one holding its files.
95
+ - **Every command has a budget.** A command with no timeout gets the backend's default (120 s); `timeout=0` gets the
96
+ daemon's longest; a command past its budget answers exit 124 with what it printed. A process left in the background is
97
+ stopped when its command ends.
98
+
99
+ ## Retries
100
+
101
+ LangGraph and LangChain retry nothing unless you ask: retries are off by default in both. You turn them on with a
102
+ `RetryPolicy` on a graph node (or on every node, with `set_node_defaults`), or with `ToolRetryMiddleware` on an agent.
103
+ Once on, both frameworks' default rules retry a plimsoll error, including one after which the code may already have run,
104
+ so a retry can run it a second time: a file appended to twice, a request sent twice. `retry_on` is the rule to give them
105
+ instead. It retries only a refusal plimsoll marked as having run nothing, and only when another try may pass: the daemon
106
+ was at capacity, or could not be reached before anything was sent. Every other error is left to stop the run.
107
+
108
+ ```python
109
+ from langchain.agents.middleware import ToolRetryMiddleware
110
+ from langgraph.types import RetryPolicy
111
+ from langchain_plimsoll import retry_on
112
+
113
+ # A graph node that calls plimsoll
114
+ builder.add_node("run", node, retry_policy=RetryPolicy(retry_on=retry_on))
115
+
116
+ # A Deep Agents agent: its execute and file tools
117
+ agent = create_deep_agent(model=model, backend=backend, middleware=[ToolRetryMiddleware(retry_on=retry_on)])
118
+ ```
119
+
120
+ - **Where it matters.** The Deep Agents backend raises on a refusal (above), and so do `Client` and `CodeExecutor`
121
+ called from a node of your own: that is where a retry rule meets plimsoll. `PlimsollCodeTool` never raises for a
122
+ refusal or a lost answer; it answers with them as its JSON (`ran` false or `"unknown"`), so no retry rule sees them.
123
+ - **Which refusals.** Every plimsoll error says whether anything ran: a refusal raised before any code was handed to the
124
+ sandbox carries a reason, and an error without one may have followed execution
125
+ ([what comes back, and what it means (EXTERNAL · source repo ↗)](https://github.com/plimsollmark/plimsoll/blob/main/docs/run-results.md)).
126
+ `retry_on` says yes to the reasons `capacity` (the daemon was full or over a rate limit and turned the run away) and
127
+ `environment` (the daemon or its token check could not be reached before anything was sent, or a session's sandbox
128
+ could not be read back before the call). It says no to the others, since the same daemon answers them the same way:
129
+ `request`, `permission`, `protocol`, `unsupported` and `isolation`. One `environment` cause, an image outside your
130
+ software rule, fails the same way each time; a retry of it costs the policy's few attempts and runs nothing.
131
+ - **Only plimsoll's errors.** It says no to every exception that is not plimsoll's (every plimsoll error is a
132
+ `plimsoll_client.PlimsollError`). For a node that also calls a service whose errors you want retried, decide both in
133
+ one rule: `RetryPolicy(retry_on=lambda e: retry_on(e) if isinstance(e, PlimsollError) else your_rule(e))`. A second
134
+ `RetryPolicy` on the same node does not do this: LangGraph uses the first policy whose rule says yes, so a later policy
135
+ with LangGraph's default rule would still retry plimsoll's errors.
136
+
137
+ ## What it does, and what it refuses to do
138
+
139
+ - **No fallback.** Both parts require the `kernel` isolation tier by default (gVisor, a user-space kernel, or a virtual
140
+ machine). Against a daemon that states less, they refuse before any code is sent. Lowering the floor is a choice you
141
+ make in code: `CodeExecutor(client, minimum_isolation="container")`, `PlimsollSandbox.open(client,
142
+ minimum_isolation="container")`.
143
+ - **A checked record of every run.** Each answer carries a run record, the daemon's statement of what was sent, what came
144
+ back and where it ran, and the client recomputes its digests before handing the result on.
145
+ - **One web API, if you grant one, for the tool.** `CodeExecutor(client, grant_profile="lights", api_description=...)`,
146
+ with a profile the operator configured on the daemon and the description `plimsoll-specgen` wrote for it, lets the
147
+ tool's code call that profile's routes, and nothing else, through plimsoll's broker, which adds the API's key outside
148
+ the sandbox. JavaScript works on every provider that gives a project run a grant; Python too on docker and openshell,
149
+ through `import plimsoll_host`. The tool reads the daemon's answer when it is built, offers only the languages that
150
+ can call the API there, and tells the model the routes; the model never names a profile. The code can do whatever an
151
+ allowed route permits and read whatever it returns
152
+ ([capability grants (EXTERNAL · source repo ↗)](https://github.com/plimsollmark/plimsoll/blob/main/docs/capability-grants.md#from-an-agent-kits-code-tool)).
153
+ The Deep Agents backend's calls take no grant.
154
+ - **Tested against LangChain's own suites.** The tool passes LangChain's standard tool tests (`langchain-tests`). The
155
+ backend passes LangChain's standard sandbox tests on a real plimsoll session, except seven cases where
156
+ langchain-tests 1.1.9 and deepagents 0.7.38 disagree with each other (six about what `glob` returns, one about writing
157
+ over an existing file); they fail the same way for any backend built on Deep Agents' `BaseSandbox`.
158
+
159
+ ## Tracing
160
+
161
+ If you turn on LangSmith tracing, LangChain sends each tool call's inputs and outputs, which here are the model's code
162
+ and what it printed, to LangSmith. That is LangChain's switch, not this package's; LangSmith's `hide_inputs` and
163
+ `hide_outputs` settings turn it off
164
+ ([LangSmith tracing (EXTERNAL · official docs ↗)](https://docs.langchain.com/langsmith/mask-inputs-outputs)).
165
+
166
+ `PlimsollCodeTool` adds two things of its own, and neither carries the code or its output:
167
+
168
+ - **One ID in both logs.** Each call sends the LangChain run ID of the tool call (the ID LangSmith shows for that run) as
169
+ the run's `trace_id`, which the daemon writes on the run's audit line and does nothing else with. To find what the
170
+ daemon did for a traced tool call, search its log for the run's ID. The daemon accepts 1 to 64 characters of letters,
171
+ digits and `.`, `_`, `:` and `-`; LangChain's IDs are UUIDs, which fit. An ID you set yourself (`run_id` in the
172
+ config) that does not fit is not sent at all, never cut short, since a shortened ID would match nothing.
173
+ - **Labels on the tool call's run.** When the call is traced, the tool adds metadata to its run: `plimsoll_record_sha256`
174
+ (the run record's digest, which the client checked), `plimsoll_isolation` (the tier the code ran behind) and
175
+ `plimsoll_isolation_qualification` (the sentence a provider states beside its tier, when it states one),
176
+ `plimsoll_dispatched` (`yes`, `no` or `unknown`: whether the run reached the provider, or an error left that open),
177
+ `plimsoll_refused` (the reason, when it did not) and `plimsoll_outcome` (how the run ended: `completed`,
178
+ `setup_failed`, `timed_out`, `protocol_error` or `runner_failed`). Each value is checked against the values plimsoll
179
+ can state and left out if it does not match, so nothing the code printed can reach these fields. In LangSmith you can
180
+ then filter for runs that were refused, ran below a tier, or may have run without an answer. With tracing off the tool
181
+ adds nothing, and it imports no part of LangSmith that is not already loaded.
182
+
183
+ ## What it is not
184
+
185
+ - **Not a hosted service.** You run `plimsolld` on your own machine or cluster, with a provider you pick: docker under
186
+ gVisor, an E2B or Docker Cloud microVM, a Kubernetes agent-sandbox pod, and others
187
+ ([isolation tiers (EXTERNAL · source repo ↗)](https://github.com/plimsollmark/plimsoll/blob/main/docs/isolation-tiers.md)).
188
+ The Deep Agents backend needs a provider with sessions: docker, E2B or OpenShell.
189
+ - **Not attestation.** A reported tier is the daemon's configuration and provider evidence plus a startup check, not
190
+ hardware proof.
191
+
192
+ ## Versions
193
+
194
+ This package's version moves with plimsoll's releases, and each version requires the `plimsoll-client` release of the same
195
+ number, which holds the code (`plimsoll_client.langchain` and `plimsoll_client.deepagents`). `Client`, `CodeExecutor`,
196
+ `PlimsollCodeTool` and `PlimsollSandbox` here are plimsoll-client's own classes, gathered so an agent imports from one
197
+ place. It is tested with, and installs with, exactly langchain-core 1.8.0, and its `deepagents` extra with exactly
198
+ deepagents 0.7.38: pip installs those versions beside it and refuses others, and each new release of either is tested
199
+ before its pin moves.
200
+
201
+ Licensed under the Apache License 2.0. Source:
202
+ [plimsoll on GitHub (EXTERNAL · source repo ↗)](https://github.com/plimsollmark/plimsoll/tree/main/clients/python/langchain-plimsoll).
@@ -0,0 +1,43 @@
1
+ [build-system]
2
+ requires = ["setuptools>=77"]
3
+ build-backend = "setuptools.build_meta"
4
+
5
+ # A LangChain integration package, named as LangChain's integration guide asks (an
6
+ # independent langchain-<provider> package on PyPI; LangChain accepts no new integrations
7
+ # into its own repositories). The tool is plimsoll_client.langchain.PlimsollCodeTool and the
8
+ # Deep Agents backend plimsoll_client.deepagents.PlimsollSandbox; this package depends on
9
+ # that exact plimsoll-client release, and its version moves with plimsoll's releases
10
+ # (TestClientVersionsAgree in clients/typescript checks both). langchain-core and deepagents
11
+ # need Python 3.11 or later.
12
+ [project]
13
+ name = "langchain-plimsoll"
14
+ version = "0.28.0"
15
+ description = "A LangChain tool and a Deep Agents sandbox backend that run agent-written code in plimsoll, a self-hosted sandbox that refuses to run below the isolation you require"
16
+ readme = "README.md"
17
+ license = "Apache-2.0"
18
+ license-files = ["LICENSE"]
19
+ requires-python = ">=3.11"
20
+ dependencies = [
21
+ "langchain-core==1.8.0",
22
+ "plimsoll-client==0.28.0",
23
+ ]
24
+ keywords = ["langchain", "langgraph", "deepagents", "sandbox", "code execution", "agents", "gvisor"]
25
+ classifiers = [
26
+ "Programming Language :: Python :: 3",
27
+ "Programming Language :: Python :: 3 :: Only",
28
+ "Typing :: Typed",
29
+ ]
30
+
31
+ [project.optional-dependencies]
32
+ deepagents = ["deepagents==0.7.38"]
33
+
34
+ [project.urls]
35
+ Homepage = "https://github.com/plimsollmark/plimsoll/tree/main/clients/python/langchain-plimsoll"
36
+ Source = "https://github.com/plimsollmark/plimsoll/tree/main/clients/python/langchain-plimsoll"
37
+ Issues = "https://github.com/plimsollmark/plimsoll/issues"
38
+
39
+ [tool.setuptools.packages.find]
40
+ where = ["src"]
41
+
42
+ [tool.setuptools.package-data]
43
+ langchain_plimsoll = ["py.typed"]
@@ -0,0 +1,4 @@
1
+ [egg_info]
2
+ tag_build =
3
+ tag_date = 0
4
+
@@ -0,0 +1,45 @@
1
+ """Run a LangChain, LangGraph or Deep Agents agent's code in plimsoll.
2
+
3
+ from langchain_plimsoll import Client, CodeExecutor, PlimsollCodeTool
4
+
5
+ executor = CodeExecutor(Client("https://plimsoll.internal:8746", token=TOKEN))
6
+ tools = [PlimsollCodeTool(executor)] # create_agent, ToolNode, any LangGraph graph
7
+
8
+ from langchain_plimsoll import PlimsollSandbox # pip install 'langchain-plimsoll[deepagents]'
9
+
10
+ with PlimsollSandbox.open(Client("https://plimsoll.internal:8746", token=TOKEN)) as backend:
11
+ agent = create_deep_agent(model=..., backend=backend, system_prompt=f"Work in {backend.root}.")
12
+
13
+ from langchain_plimsoll import retry_on # LangGraph's RetryPolicy, LangChain's ToolRetryMiddleware
14
+
15
+ builder.add_node("run", node, retry_policy=RetryPolicy(retry_on=retry_on))
16
+
17
+ The names are plimsoll-client's own, gathered here so an agent imports from one place:
18
+ PlimsollCodeTool is plimsoll_client.langchain.PlimsollCodeTool (a fresh sandbox per call),
19
+ PlimsollSandbox is plimsoll_client.deepagents.PlimsollSandbox (one plimsoll session as a Deep
20
+ Agents backend), Client is plimsoll_client.Client, CodeExecutor is
21
+ plimsoll_client.execution.CodeExecutor and retry_on is plimsoll_client.retry_on (retry only a
22
+ refusal that ran nothing and may pass on another try). Both require the kernel isolation tier
23
+ by default and refuse, before any code is sent, a daemon that states less.
24
+ """
25
+
26
+ from typing import Any
27
+
28
+ from plimsoll_client import Client, retry_on
29
+ from plimsoll_client.execution import CodeExecutor
30
+ from plimsoll_client.langchain import PlimsollCodeTool
31
+
32
+ __all__ = ["Client", "CodeExecutor", "PlimsollCodeTool", "PlimsollSandbox", "retry_on"]
33
+ __version__ = "0.28.0"
34
+
35
+
36
+ def __getattr__(name: str) -> Any:
37
+ # The backend needs deepagents, an extra: imported on first use, so the tool alone
38
+ # needs only langchain-core.
39
+ if name == "PlimsollSandbox":
40
+ try:
41
+ from plimsoll_client.deepagents import PlimsollSandbox
42
+ except ImportError as e:
43
+ raise ImportError("langchain_plimsoll.PlimsollSandbox needs Deep Agents: pip install 'langchain-plimsoll[deepagents]'") from e
44
+ return PlimsollSandbox
45
+ raise AttributeError(f"module 'langchain_plimsoll' has no attribute {name!r}")
@@ -0,0 +1,223 @@
1
+ Metadata-Version: 2.4
2
+ Name: langchain-plimsoll
3
+ Version: 0.28.0
4
+ Summary: A LangChain tool and a Deep Agents sandbox backend that run agent-written code in plimsoll, a self-hosted sandbox that refuses to run below the isolation you require
5
+ License-Expression: Apache-2.0
6
+ Project-URL: Homepage, https://github.com/plimsollmark/plimsoll/tree/main/clients/python/langchain-plimsoll
7
+ Project-URL: Source, https://github.com/plimsollmark/plimsoll/tree/main/clients/python/langchain-plimsoll
8
+ Project-URL: Issues, https://github.com/plimsollmark/plimsoll/issues
9
+ Keywords: langchain,langgraph,deepagents,sandbox,code execution,agents,gvisor
10
+ Classifier: Programming Language :: Python :: 3
11
+ Classifier: Programming Language :: Python :: 3 :: Only
12
+ Classifier: Typing :: Typed
13
+ Requires-Python: >=3.11
14
+ Description-Content-Type: text/markdown
15
+ License-File: LICENSE
16
+ Requires-Dist: langchain-core==1.8.0
17
+ Requires-Dist: plimsoll-client==0.28.0
18
+ Provides-Extra: deepagents
19
+ Requires-Dist: deepagents==0.7.38; extra == "deepagents"
20
+ Dynamic: license-file
21
+
22
+ # langchain-plimsoll
23
+
24
+ An agent that writes code has to run it somewhere, and "somewhere" is usually your machine or a hosted sandbox you
25
+ cannot inspect. This package runs a LangChain, LangGraph or Deep Agents agent's code in plimsoll, a sandbox service you
26
+ run yourself, and it has no fallback: if the sandbox is weaker than you asked for, nothing runs.
27
+
28
+ It has two parts:
29
+
30
+ - **`PlimsollCodeTool`**, a LangChain tool (`plimsoll_run_code`) that runs each call in a fresh sandbox. LangGraph's
31
+ `ToolNode`, `create_agent` and any graph of your own call it like any other LangChain tool.
32
+ - **`PlimsollSandbox`**, a Deep Agents sandbox backend over one plimsoll session: one sandbox kept for the agent's whole
33
+ run, so the files it writes and the shell's directory are there for its next command. Deep Agents gives the agent its
34
+ `execute` tool and its file tools (`ls`, `read_file`, `write_file`, `edit_file`, `glob`, `grep`) on top of it
35
+ ([Deep Agents sandboxes (EXTERNAL · official docs ↗)](https://docs.langchain.com/oss/python/deepagents/sandboxes)).
36
+
37
+ ## Install
38
+
39
+ ```sh
40
+ pip install langchain-plimsoll # the tool: needs only langchain-core
41
+ pip install 'langchain-plimsoll[deepagents]' # and the Deep Agents backend
42
+ ```
43
+
44
+ The tool needs only `langchain-core`. The example below builds its agent with `create_agent`, which is in `langchain`:
45
+ `pip install langchain` for it.
46
+
47
+ It needs a running `plimsolld`, the plimsoll daemon, reachable from the agent
48
+ ([getting started (EXTERNAL · source repo ↗)](https://github.com/plimsollmark/plimsoll/blob/main/docs/getting-started.md)).
49
+ The tool's default language is Python, which on docker needs a project image with `python3` in it
50
+ (`SANDBOX_DOCKER_PROJECT_IMAGE=plimsoll/sandbox-python:latest`): on the default image every Python call comes back
51
+ refused. Python 3.11 or later, as langchain-core requires.
52
+
53
+ ## Use the tool
54
+
55
+ ```python
56
+ from langchain.agents import create_agent
57
+ from langchain_plimsoll import Client, CodeExecutor, PlimsollCodeTool
58
+
59
+ executor = CodeExecutor(Client("https://plimsoll.internal:8746", token=TOKEN))
60
+ agent = create_agent(model, tools=[PlimsollCodeTool(executor)])
61
+ ```
62
+
63
+ The agent passes the tool code (Python or JavaScript) and optional input files, and gets back JSON saying whether the code
64
+ ran (yes, no with the reason, or unknown), the output, the exit code, whether anything was cut short or timed out, the
65
+ isolation tier the code ran behind, and the SHA-256 of the run's record.
66
+
67
+ ## Use the Deep Agents backend
68
+
69
+ ```python
70
+ from deepagents import create_deep_agent
71
+ from langchain_plimsoll import Client, PlimsollSandbox
72
+
73
+ with PlimsollSandbox.open(Client("https://plimsoll.internal:8746", token=TOKEN)) as backend:
74
+ agent = create_deep_agent(model=model, backend=backend, system_prompt=f"Work in {backend.root}.")
75
+ agent.invoke({"messages": [{"role": "user", "content": "Write and test a CSV parser."}]})
76
+ ```
77
+
78
+ Deep Agents does not tell the model where to work; `backend.root` is the sandbox's work directory (`/work` on docker),
79
+ which is where the agent's files belong. Leaving the `with` block closes the session, and the sandbox and its files are
80
+ gone.
81
+
82
+ What the backend needs and does:
83
+
84
+ - **Python 3 in the sandbox.** Deep Agents' file tools are small `python3` scripts it runs through `execute`. `open`
85
+ refuses a daemon that does not state Python before any sandbox is made; on docker that is the `plimsoll/sandbox-python`
86
+ image (`SANDBOX_DOCKER_PROJECT_IMAGE`).
87
+ - **Searches that never answer "no matches" by mistake.** Deep Agents' `grep` sends GNU grep options with its errors
88
+ discarded. plimsoll's docker images are Alpine Linux, whose `grep` is BusyBox's and has neither option, so every such
89
+ search would come back empty. When the sandbox's `grep` is not GNU grep, the backend sends each search through Deep
90
+ Agents' own `python3` search instead, which reports its errors. It skips hidden files, as Python's glob does.
91
+ - **A refusal stops the run.** If plimsoll refuses a command (the daemon is at capacity, the session has ended, the tier
92
+ is too low), or a command may have run without an answer to show for it, the backend raises. It never hands the model a
93
+ refusal dressed as command output, because Deep Agents reads that output as data: its `ls` reads text it cannot parse
94
+ as an empty directory. A command that runs and fails is a normal result with its exit code.
95
+ - **A thread resumed on another worker keeps its sandbox.** Open the backend with a checkpoint store and a key, and
96
+ attach to it from the worker that resumes the thread:
97
+
98
+ ```python
99
+ from plimsoll_client import MemoryCheckpointStore # for tests: use storage only your app can write
100
+
101
+ key = f"{user_id}/{thread_id}" # from identities your app verified, never from model or user text
102
+ backend = PlimsollSandbox.open(Client(url, token, daemon_name="prod"), checkpoint_store=store, checkpoint_key=key)
103
+ # On the worker that resumes the thread:
104
+ backend = PlimsollSandbox.attach(Client(url, token, daemon_name="prod"), key, store=store)
105
+ ```
106
+
107
+ The second worker gets the same sandbox and its files, and the first is refused from then on. A store is an object
108
+ with `load` and an atomic `compare_and_set`; every call passes through it, so two workers never send at once, and a
109
+ worker that died mid-command leaves a record the next one takes over from (if that command ran with its answer lost,
110
+ the next one's first command may not repeat it). A missing key raises `LookupError`, never a new sandbox. The daemon
111
+ needs callers with tokens: open development mode refuses the takeover, and anyone holding the token and a session's run
112
+ records can take it over. The rules:
113
+ [carrying a session on from another worker (EXTERNAL · source repo ↗)](https://github.com/plimsollmark/plimsoll/blob/main/docs/sessions.md#carrying-a-session-on-from-another-worker).
114
+ When the session ends (the daemon's lifetime or idle limit), every later call raises; the backend never opens a fresh
115
+ sandbox in its place, since the agent would take it for the one holding its files.
116
+ - **Every command has a budget.** A command with no timeout gets the backend's default (120 s); `timeout=0` gets the
117
+ daemon's longest; a command past its budget answers exit 124 with what it printed. A process left in the background is
118
+ stopped when its command ends.
119
+
120
+ ## Retries
121
+
122
+ LangGraph and LangChain retry nothing unless you ask: retries are off by default in both. You turn them on with a
123
+ `RetryPolicy` on a graph node (or on every node, with `set_node_defaults`), or with `ToolRetryMiddleware` on an agent.
124
+ Once on, both frameworks' default rules retry a plimsoll error, including one after which the code may already have run,
125
+ so a retry can run it a second time: a file appended to twice, a request sent twice. `retry_on` is the rule to give them
126
+ instead. It retries only a refusal plimsoll marked as having run nothing, and only when another try may pass: the daemon
127
+ was at capacity, or could not be reached before anything was sent. Every other error is left to stop the run.
128
+
129
+ ```python
130
+ from langchain.agents.middleware import ToolRetryMiddleware
131
+ from langgraph.types import RetryPolicy
132
+ from langchain_plimsoll import retry_on
133
+
134
+ # A graph node that calls plimsoll
135
+ builder.add_node("run", node, retry_policy=RetryPolicy(retry_on=retry_on))
136
+
137
+ # A Deep Agents agent: its execute and file tools
138
+ agent = create_deep_agent(model=model, backend=backend, middleware=[ToolRetryMiddleware(retry_on=retry_on)])
139
+ ```
140
+
141
+ - **Where it matters.** The Deep Agents backend raises on a refusal (above), and so do `Client` and `CodeExecutor`
142
+ called from a node of your own: that is where a retry rule meets plimsoll. `PlimsollCodeTool` never raises for a
143
+ refusal or a lost answer; it answers with them as its JSON (`ran` false or `"unknown"`), so no retry rule sees them.
144
+ - **Which refusals.** Every plimsoll error says whether anything ran: a refusal raised before any code was handed to the
145
+ sandbox carries a reason, and an error without one may have followed execution
146
+ ([what comes back, and what it means (EXTERNAL · source repo ↗)](https://github.com/plimsollmark/plimsoll/blob/main/docs/run-results.md)).
147
+ `retry_on` says yes to the reasons `capacity` (the daemon was full or over a rate limit and turned the run away) and
148
+ `environment` (the daemon or its token check could not be reached before anything was sent, or a session's sandbox
149
+ could not be read back before the call). It says no to the others, since the same daemon answers them the same way:
150
+ `request`, `permission`, `protocol`, `unsupported` and `isolation`. One `environment` cause, an image outside your
151
+ software rule, fails the same way each time; a retry of it costs the policy's few attempts and runs nothing.
152
+ - **Only plimsoll's errors.** It says no to every exception that is not plimsoll's (every plimsoll error is a
153
+ `plimsoll_client.PlimsollError`). For a node that also calls a service whose errors you want retried, decide both in
154
+ one rule: `RetryPolicy(retry_on=lambda e: retry_on(e) if isinstance(e, PlimsollError) else your_rule(e))`. A second
155
+ `RetryPolicy` on the same node does not do this: LangGraph uses the first policy whose rule says yes, so a later policy
156
+ with LangGraph's default rule would still retry plimsoll's errors.
157
+
158
+ ## What it does, and what it refuses to do
159
+
160
+ - **No fallback.** Both parts require the `kernel` isolation tier by default (gVisor, a user-space kernel, or a virtual
161
+ machine). Against a daemon that states less, they refuse before any code is sent. Lowering the floor is a choice you
162
+ make in code: `CodeExecutor(client, minimum_isolation="container")`, `PlimsollSandbox.open(client,
163
+ minimum_isolation="container")`.
164
+ - **A checked record of every run.** Each answer carries a run record, the daemon's statement of what was sent, what came
165
+ back and where it ran, and the client recomputes its digests before handing the result on.
166
+ - **One web API, if you grant one, for the tool.** `CodeExecutor(client, grant_profile="lights", api_description=...)`,
167
+ with a profile the operator configured on the daemon and the description `plimsoll-specgen` wrote for it, lets the
168
+ tool's code call that profile's routes, and nothing else, through plimsoll's broker, which adds the API's key outside
169
+ the sandbox. JavaScript works on every provider that gives a project run a grant; Python too on docker and openshell,
170
+ through `import plimsoll_host`. The tool reads the daemon's answer when it is built, offers only the languages that
171
+ can call the API there, and tells the model the routes; the model never names a profile. The code can do whatever an
172
+ allowed route permits and read whatever it returns
173
+ ([capability grants (EXTERNAL · source repo ↗)](https://github.com/plimsollmark/plimsoll/blob/main/docs/capability-grants.md#from-an-agent-kits-code-tool)).
174
+ The Deep Agents backend's calls take no grant.
175
+ - **Tested against LangChain's own suites.** The tool passes LangChain's standard tool tests (`langchain-tests`). The
176
+ backend passes LangChain's standard sandbox tests on a real plimsoll session, except seven cases where
177
+ langchain-tests 1.1.9 and deepagents 0.7.38 disagree with each other (six about what `glob` returns, one about writing
178
+ over an existing file); they fail the same way for any backend built on Deep Agents' `BaseSandbox`.
179
+
180
+ ## Tracing
181
+
182
+ If you turn on LangSmith tracing, LangChain sends each tool call's inputs and outputs, which here are the model's code
183
+ and what it printed, to LangSmith. That is LangChain's switch, not this package's; LangSmith's `hide_inputs` and
184
+ `hide_outputs` settings turn it off
185
+ ([LangSmith tracing (EXTERNAL · official docs ↗)](https://docs.langchain.com/langsmith/mask-inputs-outputs)).
186
+
187
+ `PlimsollCodeTool` adds two things of its own, and neither carries the code or its output:
188
+
189
+ - **One ID in both logs.** Each call sends the LangChain run ID of the tool call (the ID LangSmith shows for that run) as
190
+ the run's `trace_id`, which the daemon writes on the run's audit line and does nothing else with. To find what the
191
+ daemon did for a traced tool call, search its log for the run's ID. The daemon accepts 1 to 64 characters of letters,
192
+ digits and `.`, `_`, `:` and `-`; LangChain's IDs are UUIDs, which fit. An ID you set yourself (`run_id` in the
193
+ config) that does not fit is not sent at all, never cut short, since a shortened ID would match nothing.
194
+ - **Labels on the tool call's run.** When the call is traced, the tool adds metadata to its run: `plimsoll_record_sha256`
195
+ (the run record's digest, which the client checked), `plimsoll_isolation` (the tier the code ran behind) and
196
+ `plimsoll_isolation_qualification` (the sentence a provider states beside its tier, when it states one),
197
+ `plimsoll_dispatched` (`yes`, `no` or `unknown`: whether the run reached the provider, or an error left that open),
198
+ `plimsoll_refused` (the reason, when it did not) and `plimsoll_outcome` (how the run ended: `completed`,
199
+ `setup_failed`, `timed_out`, `protocol_error` or `runner_failed`). Each value is checked against the values plimsoll
200
+ can state and left out if it does not match, so nothing the code printed can reach these fields. In LangSmith you can
201
+ then filter for runs that were refused, ran below a tier, or may have run without an answer. With tracing off the tool
202
+ adds nothing, and it imports no part of LangSmith that is not already loaded.
203
+
204
+ ## What it is not
205
+
206
+ - **Not a hosted service.** You run `plimsolld` on your own machine or cluster, with a provider you pick: docker under
207
+ gVisor, an E2B or Docker Cloud microVM, a Kubernetes agent-sandbox pod, and others
208
+ ([isolation tiers (EXTERNAL · source repo ↗)](https://github.com/plimsollmark/plimsoll/blob/main/docs/isolation-tiers.md)).
209
+ The Deep Agents backend needs a provider with sessions: docker, E2B or OpenShell.
210
+ - **Not attestation.** A reported tier is the daemon's configuration and provider evidence plus a startup check, not
211
+ hardware proof.
212
+
213
+ ## Versions
214
+
215
+ This package's version moves with plimsoll's releases, and each version requires the `plimsoll-client` release of the same
216
+ number, which holds the code (`plimsoll_client.langchain` and `plimsoll_client.deepagents`). `Client`, `CodeExecutor`,
217
+ `PlimsollCodeTool` and `PlimsollSandbox` here are plimsoll-client's own classes, gathered so an agent imports from one
218
+ place. It is tested with, and installs with, exactly langchain-core 1.8.0, and its `deepagents` extra with exactly
219
+ deepagents 0.7.38: pip installs those versions beside it and refuses others, and each new release of either is tested
220
+ before its pin moves.
221
+
222
+ Licensed under the Apache License 2.0. Source:
223
+ [plimsoll on GitHub (EXTERNAL · source repo ↗)](https://github.com/plimsollmark/plimsoll/tree/main/clients/python/langchain-plimsoll).
@@ -0,0 +1,10 @@
1
+ LICENSE
2
+ README.md
3
+ pyproject.toml
4
+ src/langchain_plimsoll/__init__.py
5
+ src/langchain_plimsoll/py.typed
6
+ src/langchain_plimsoll.egg-info/PKG-INFO
7
+ src/langchain_plimsoll.egg-info/SOURCES.txt
8
+ src/langchain_plimsoll.egg-info/dependency_links.txt
9
+ src/langchain_plimsoll.egg-info/requires.txt
10
+ src/langchain_plimsoll.egg-info/top_level.txt
@@ -0,0 +1,5 @@
1
+ langchain-core==1.8.0
2
+ plimsoll-client==0.28.0
3
+
4
+ [deepagents]
5
+ deepagents==0.7.38