stable-harness 0.0.48 → 0.0.49
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.
package/README.md
CHANGED
|
@@ -13,6 +13,11 @@ tool repair, and protocol access.
|
|
|
13
13
|
It is not another agent execution framework. Upstream frameworks own execution
|
|
14
14
|
semantics. Stable Harness owns the runtime boundary around them.
|
|
15
15
|
|
|
16
|
+
In the EasyNet World stack, Stable Harness is the runtime substrate under Flev.
|
|
17
|
+
Better Call hardens the tool-call boundary, Stable Harness governs the agent
|
|
18
|
+
workspace runtime, and Flev turns that governed workspace into user-facing
|
|
19
|
+
product surfaces such as Studio, chat, embed, review, and delivery workflows.
|
|
20
|
+
|
|
16
21
|
## Why Use It
|
|
17
22
|
|
|
18
23
|
Agent frameworks are good at deciding what an agent should do next. Production
|
|
@@ -224,6 +229,8 @@ This is constrained repair, not silent magic:
|
|
|
224
229
|
- [Getting started](docs/guides/getting-started.md)
|
|
225
230
|
- [Workspace authoring](docs/guides/workspace-authoring.md)
|
|
226
231
|
- [Workspace Docker build](docs/guides/workspace-docker-build.md)
|
|
232
|
+
- [Runtime substrate for Flev](docs/product/runtime-substrate-for-flev.md)
|
|
233
|
+
- [Runtime governance proof](docs/guides/runtime-governance-proof.md)
|
|
227
234
|
- [Integration guide](docs/guides/integration-guide.md)
|
|
228
235
|
- [Operator runbook](docs/guides/operator-runbook.md)
|
|
229
236
|
- [Adoption playbook](docs/product/adoption-playbook.md)
|
|
@@ -0,0 +1,54 @@
|
|
|
1
|
+
# Runtime Governance Proof
|
|
2
|
+
|
|
3
|
+
Use this page as the compact proof checklist for public claims about Stable
|
|
4
|
+
Harness runtime governance. It is meant to give Flev and the EasyNet World site
|
|
5
|
+
specific runtime evidence to cite without turning Stable Harness into the
|
|
6
|
+
end-user product.
|
|
7
|
+
|
|
8
|
+
## Minimal Proof Run
|
|
9
|
+
|
|
10
|
+
```bash
|
|
11
|
+
npm install
|
|
12
|
+
npm run build
|
|
13
|
+
stable-harness -w ./examples/minimal-deepagents "hello stable harness"
|
|
14
|
+
```
|
|
15
|
+
|
|
16
|
+
Inspect the workspace without executing the agent:
|
|
17
|
+
|
|
18
|
+
```bash
|
|
19
|
+
stable-harness -w ./examples/minimal-deepagents
|
|
20
|
+
stable-harness agent render orchestra -w ./examples/minimal-deepagents
|
|
21
|
+
```
|
|
22
|
+
|
|
23
|
+
Start the protocol facade:
|
|
24
|
+
|
|
25
|
+
```bash
|
|
26
|
+
stable-harness start -w ./examples/minimal-deepagents --port 8642
|
|
27
|
+
curl -fsS http://127.0.0.1:8642/v1/models
|
|
28
|
+
```
|
|
29
|
+
|
|
30
|
+
Package the same workspace as a portable Docker runtime:
|
|
31
|
+
|
|
32
|
+
```bash
|
|
33
|
+
stable-harness build --target docker -w ./examples/minimal-deepagents --output ./dist/workspace-docker
|
|
34
|
+
cd ./dist/workspace-docker
|
|
35
|
+
STABLE_HARNESS_PORT=8642 docker compose up --build
|
|
36
|
+
curl -fsS http://127.0.0.1:8642/v1/models
|
|
37
|
+
```
|
|
38
|
+
|
|
39
|
+
## What The Proof Shows
|
|
40
|
+
|
|
41
|
+
| Claim | Evidence surface |
|
|
42
|
+
| --- | --- |
|
|
43
|
+
| Workspace configuration is reviewable | `agent render`, runtime inventory, YAML workspace files |
|
|
44
|
+
| Provider and model choice stay behind runtime boundaries | runtime config and `/v1/models` protocol output |
|
|
45
|
+
| Tool policy can be governed before execution | tool gateway plus Better Call repair/validation events |
|
|
46
|
+
| Runtime work can be inspected | request, session, event, artifact, approval, and replay APIs |
|
|
47
|
+
| Deployment has a stable boundary | generated Dockerfile, Compose file, `.env.example`, and `stable-workspace.json` |
|
|
48
|
+
|
|
49
|
+
## Public Copy Boundary
|
|
50
|
+
|
|
51
|
+
Stable Harness proof should describe runtime governance: sessions, events,
|
|
52
|
+
providers, tools, approvals, receipts, replay, protocols, and packaging. Flev
|
|
53
|
+
proof should describe user-facing product surfaces: Studio, chat, embed, review,
|
|
54
|
+
templates, and vertical workflow delivery.
|
|
@@ -0,0 +1,55 @@
|
|
|
1
|
+
# Stable Harness as Flev's Runtime Substrate
|
|
2
|
+
|
|
3
|
+
Stable Harness is the runtime and operator-control layer under Flev. It should
|
|
4
|
+
stay generic: it gives agent workspaces a governed execution boundary, while
|
|
5
|
+
Flev owns the user-facing product workspace, Studio views, embedded clients,
|
|
6
|
+
review flows, and vertical product packaging.
|
|
7
|
+
|
|
8
|
+
```text
|
|
9
|
+
Better Call -> Stable Harness -> Flev -> end-to-end products
|
|
10
|
+
tool-call reliability -> runtime control -> product workspace -> vertical workflow
|
|
11
|
+
```
|
|
12
|
+
|
|
13
|
+
## What Stable Harness Owns
|
|
14
|
+
|
|
15
|
+
- Workspace loading and validation.
|
|
16
|
+
- Session, request, run, and event lifecycle.
|
|
17
|
+
- Provider and model resolution.
|
|
18
|
+
- Tool gateway policy, including Better Call integration.
|
|
19
|
+
- Approval checkpoints and governance state.
|
|
20
|
+
- Runtime receipts for tool calls, provider choices, approvals, artifacts, and
|
|
21
|
+
final output.
|
|
22
|
+
- Runtime APIs, CLI, protocol facades, and in-process SDK access.
|
|
23
|
+
- Generic workspace-to-Docker artifact generation.
|
|
24
|
+
|
|
25
|
+
## What Stable Harness Does Not Own
|
|
26
|
+
|
|
27
|
+
- Flev Studio product layout and interaction decisions.
|
|
28
|
+
- Vertical workflow UX such as DevOps, finance, research, or compliance
|
|
29
|
+
product flows.
|
|
30
|
+
- Business-specific orchestration rules.
|
|
31
|
+
- Product packaging, customer account model, billing, or onboarding.
|
|
32
|
+
- Backend-native reasoning, planning, or delegation semantics when the selected
|
|
33
|
+
upstream framework already owns them.
|
|
34
|
+
|
|
35
|
+
## Website-Ready Copy
|
|
36
|
+
|
|
37
|
+
English:
|
|
38
|
+
|
|
39
|
+
> Stable Harness is the runtime substrate behind Flev. It gives agent
|
|
40
|
+
> workspaces sessions, provider boundaries, tool governance, approvals, events,
|
|
41
|
+
> receipts, replay, protocol access, and Docker packaging while leaving product
|
|
42
|
+
> UX and vertical workflow design to Flev.
|
|
43
|
+
|
|
44
|
+
Chinese:
|
|
45
|
+
|
|
46
|
+
> Stable Harness 是 Flev 背后的运行时底座。它为 Agent 工作区提供 session、
|
|
47
|
+
> provider 边界、工具治理、审批、事件、运行凭证、回放、协议访问和 Docker
|
|
48
|
+
> 打包；产品界面和垂直工作流设计则由 Flev 承担。
|
|
49
|
+
|
|
50
|
+
## Design Rule
|
|
51
|
+
|
|
52
|
+
If a capability is generic runtime governance, it belongs in Stable Harness. If
|
|
53
|
+
it turns a workflow into a user-facing product, it belongs in Flev. If it repairs
|
|
54
|
+
or validates model/tool-call shape at the execution boundary, it belongs in
|
|
55
|
+
Better Call.
|
package/docs/product-boundary.md
CHANGED
|
@@ -6,6 +6,12 @@ It is not a new agent framework. It wraps upstream frameworks such as DeepAgents
|
|
|
6
6
|
|
|
7
7
|
DeepAgents is the first backend target. The current DeepAgents feature set must be audited before adding any overlapping `stable-harness` behavior.
|
|
8
8
|
|
|
9
|
+
In the EasyNet World product stack, `stable-harness` is the runtime substrate
|
|
10
|
+
under Flev. Flev owns the product workspace and user-facing surfaces. Better
|
|
11
|
+
Call owns reusable tool-call reliability. Stable Harness should expose the
|
|
12
|
+
runtime evidence and control surfaces that Flev can package, not absorb Flev's
|
|
13
|
+
vertical product behavior into the generic runtime.
|
|
14
|
+
|
|
9
15
|
## Product Mission
|
|
10
16
|
|
|
11
17
|
The product should let a team define a production agent workspace in YAML, choose an execution backend, and get stable runtime operations without rewriting the backend framework.
|