evalshift 1.2.0__tar.gz → 1.2.2__tar.gz

{evalshift-1.2.0 → evalshift-1.2.2}/CHANGELOG.md

@@ -7,6 +7,45 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [1.2.2] - 2026-10-05
+
+### Fixed
+
+- `evalshift push` printed a traceback instead of the upgrade prompt when hosted EvalShift
+  refused the run with HTTP 402 right after auto-creating the project (over the monthly run
+  or concurrent-run limit, or a subscription that stopped paying). It now prints the
+  server's message and the billing link and exits 1. A 402 while auto-creating the
+  project — what an organization whose trial has ended gets — now shows the same prompt
+  instead of an access-permission hint.
+- `evalshift push` printed a traceback when the connection failed while auto-creating the
+  project; it now prints the error and exits 1.
+
+### Changed
+
+- The EvalShift repositories moved from the `babaliauskas` GitHub account to
+  the `evalshift` organization: <https://github.com/evalshift/evalshift-cli>,
+  `evalshift-sdk` and `evalshift-action`. Links, package metadata and the
+  workflow `evalshift init --ci` scaffolds now use
+  `evalshift/evalshift-action@v0`. GitHub redirects the old names, so existing
+  workflows keep working.
+- The CI pin check (`init`, `validate`, `doctor`, `capture sync`) recognizes
+  `evalshift/evalshift-action` steps as well as the old
+  `babaliauskas/evalshift-action` name, matching owner and repository names
+  case-insensitively as GitHub does.
+- Docs and the `init --ci` workflow comment describe hosted EvalShift's current plans: a
+  30-day Pro trial for every new account, then Pro per organization (the hosted Free and
+  Team plans are gone). Local runs stay free and unlimited.
+
+## [1.2.1] - 2026-10-01
+
+### Changed
+
+- Top regressions in the HTML report now read as expandable: the worst one
+  in each section starts open, each row has a boxed caret and a
+  "Show details" / "Hide details" hint, and the header reacts on hover and
+  shows a keyboard focus ring. Printing expands every regression in
+  Chromium-based browsers.
+
 ## [1.2.0] - 2026-10-01
 
 Shipped as a minor again. The `slices` removal below is breaking by the
@@ -606,7 +645,6 @@ the key and the fix, which is the whole migration.
   that imported CLI internals (`from evalshift.models.client import …`) must
   import from `evalshift_cli`. No shim is possible — shipping any `evalshift/`
   file would recreate the collision — so this ships as a minor bump (0.14.0).
-  Design: `docs/superpowers/specs/2026-09-09-namespace-collision-design.md`.
 - Docs: the capture guides (`README.md`, `DOCS.md`, `docs/sdk.md`,
   `docs/getting-started.md`, `llms-full.txt`) now cover the SDK 0.4.0 provider
   client wrappers (`wrap_openai` / `wrap_anthropic` / `wrap_genai`), the

{evalshift-1.2.0 → evalshift-1.2.2}/PKG-INFO

@@ -1,12 +1,12 @@
 Metadata-Version: 2.5
 Name: evalshift
-Version: 1.2.0
+Version: 1.2.2
 Summary: Run your prompts on two LLMs and find out, with statistical confidence, what regressed.
-Project-URL: Homepage, https://github.com/babaliauskas/evalshift-cli
-Project-URL: Documentation, https://github.com/babaliauskas/evalshift-cli#readme
-Project-URL: Repository, https://github.com/babaliauskas/evalshift-cli
-Project-URL: Issues, https://github.com/babaliauskas/evalshift-cli/issues
-Project-URL: Changelog, https://github.com/babaliauskas/evalshift-cli/blob/main/CHANGELOG.md
+Project-URL: Homepage, https://github.com/evalshift/evalshift-cli
+Project-URL: Documentation, https://github.com/evalshift/evalshift-cli#readme
+Project-URL: Repository, https://github.com/evalshift/evalshift-cli
+Project-URL: Issues, https://github.com/evalshift/evalshift-cli/issues
+Project-URL: Changelog, https://github.com/evalshift/evalshift-cli/blob/main/CHANGELOG.md
 Author: Lukas Babaliauskas
 License-Expression: Apache-2.0
 License-File: LICENSE
@@ -57,7 +57,7 @@ Description-Content-Type: text/markdown
 
 Open-source LLM migration and regression testing for AI agents.
 
-[![CI](https://github.com/babaliauskas/evalshift-cli/actions/workflows/ci.yml/badge.svg)](https://github.com/babaliauskas/evalshift-cli/actions/workflows/ci.yml)
+[![CI](https://github.com/evalshift/evalshift-cli/actions/workflows/ci.yml/badge.svg)](https://github.com/evalshift/evalshift-cli/actions/workflows/ci.yml)
 [![License: Apache 2.0](https://img.shields.io/badge/License-Apache_2.0-blue.svg)](LICENSE)
 [![Python 3.11+](https://img.shields.io/badge/python-3.11%2B-blue.svg)](https://www.python.org/downloads/)
 [![PyPI](https://img.shields.io/pypi/v/evalshift.svg)](https://pypi.org/project/evalshift/)
@@ -82,7 +82,7 @@ statistics**: paired tests, Cohen's d, 95% CIs, and Benjamini-Hochberg
 correction across every (prompt x evaluator x slice) comparison.
 
 An eval is only worth the examples in it. That is why the
-[capture SDK](https://github.com/babaliauskas/evalshift-sdk) is part of the
+[capture SDK](https://github.com/evalshift/evalshift-sdk) is part of the
 product rather than an add-on: it records real production runs — model calls,
 tool calls, final outputs — to disk, and `evalshift capture sync` promotes them
 into golden suites. Hand-written suites are fully supported too, but captured
@@ -99,7 +99,7 @@ Four pieces, released and documented independently:
 | --- | --- | --- |
 | **SDK** — PyPI `evalshift-sdk` | Records what your agent actually did in production — model calls, tool calls, final output — as capture files on disk. Those captures become your golden suite. | [docs/sdk.md](docs/sdk.md) |
 | **CLI** — this repo, PyPI `evalshift` | Replays the suite on two models, scores, analyses, reports, bundles, pushes. | [DOCS.md](DOCS.md) |
-| **GitHub Action** — `babaliauskas/evalshift-action@v0` | Runs the pipeline on pull requests, pushes the run, posts one PR comment, sets the `evalshift/regression` status. | [docs/github-action.md](docs/github-action.md) |
+| **GitHub Action** — `evalshift/evalshift-action@v0` | Runs the pipeline on pull requests, pushes the run, posts one PR comment, sets the `evalshift/regression` status. | [docs/github-action.md](docs/github-action.md) |
 | **Hosted server** — `api.evalshift.dev`, web app at `evalshift.dev` | Optional. Stores pushed run bundles, diffs them across branches, drives PR comments and gating. | [docs/hosted.md](docs/hosted.md) |
 
 The SDK and the CLI never call each other — the interface is files under
@@ -150,7 +150,7 @@ uv pip install evalshift-sdk     # or: pip install evalshift-sdk
 From source (for contributors):
 
 ```bash
-git clone https://github.com/babaliauskas/evalshift-cli.git
+git clone https://github.com/evalshift/evalshift-cli.git
 cd evalshift-cli
 uv venv --python 3.11
 source .venv/bin/activate
@@ -168,7 +168,7 @@ candidate model to it.
 evalshift init                    # minimal capture-first evalshift.yaml
 ```
 
-Instrument the agent with [evalshift-sdk](https://github.com/babaliauskas/evalshift-sdk)
+Instrument the agent with [evalshift-sdk](https://github.com/evalshift/evalshift-sdk)
 — stdlib-only, Python 3.10+, installed with the CLI or on its own:
 
 ```python
@@ -297,7 +297,7 @@ The short version:
 
 `evalshift init --ci` scaffolds a production-shaped workflow: it discovers
 every committed suite under `.evalshift/suites/`, evaluates each on every
-pull request via [`babaliauskas/evalshift-action@v0`](https://github.com/babaliauskas/evalshift-action)
+pull request via [`evalshift/evalshift-action@v0`](https://github.com/evalshift/evalshift-action)
 (one matrix job per suite), pushes the runs to hosted EvalShift, compares
 against the latest compatible base-branch run, posts one PR comment, and
 gates merges on your `migration_policy` through a single required
@@ -423,7 +423,7 @@ Runnable projects under [`examples/`](examples/):
 
 [Apache-2.0](LICENSE). Free for any use, commercial included — no share-back
 requirement, and an explicit patent grant. The capture SDK
-([`evalshift-sdk`](https://github.com/babaliauskas/evalshift-sdk)), the piece
+([`evalshift-sdk`](https://github.com/evalshift/evalshift-sdk)), the piece
 you import into your own application, is MIT.
 
 Earlier releases stay under the license they shipped with: `0.3.0` and earlier

{evalshift-1.2.0 → evalshift-1.2.2}/README.md

@@ -2,7 +2,7 @@
 
 Open-source LLM migration and regression testing for AI agents.
 
-[![CI](https://github.com/babaliauskas/evalshift-cli/actions/workflows/ci.yml/badge.svg)](https://github.com/babaliauskas/evalshift-cli/actions/workflows/ci.yml)
+[![CI](https://github.com/evalshift/evalshift-cli/actions/workflows/ci.yml/badge.svg)](https://github.com/evalshift/evalshift-cli/actions/workflows/ci.yml)
 [![License: Apache 2.0](https://img.shields.io/badge/License-Apache_2.0-blue.svg)](LICENSE)
 [![Python 3.11+](https://img.shields.io/badge/python-3.11%2B-blue.svg)](https://www.python.org/downloads/)
 [![PyPI](https://img.shields.io/pypi/v/evalshift.svg)](https://pypi.org/project/evalshift/)
@@ -27,7 +27,7 @@ statistics**: paired tests, Cohen's d, 95% CIs, and Benjamini-Hochberg
 correction across every (prompt x evaluator x slice) comparison.
 
 An eval is only worth the examples in it. That is why the
-[capture SDK](https://github.com/babaliauskas/evalshift-sdk) is part of the
+[capture SDK](https://github.com/evalshift/evalshift-sdk) is part of the
 product rather than an add-on: it records real production runs — model calls,
 tool calls, final outputs — to disk, and `evalshift capture sync` promotes them
 into golden suites. Hand-written suites are fully supported too, but captured
@@ -44,7 +44,7 @@ Four pieces, released and documented independently:
 | --- | --- | --- |
 | **SDK** — PyPI `evalshift-sdk` | Records what your agent actually did in production — model calls, tool calls, final output — as capture files on disk. Those captures become your golden suite. | [docs/sdk.md](docs/sdk.md) |
 | **CLI** — this repo, PyPI `evalshift` | Replays the suite on two models, scores, analyses, reports, bundles, pushes. | [DOCS.md](DOCS.md) |
-| **GitHub Action** — `babaliauskas/evalshift-action@v0` | Runs the pipeline on pull requests, pushes the run, posts one PR comment, sets the `evalshift/regression` status. | [docs/github-action.md](docs/github-action.md) |
+| **GitHub Action** — `evalshift/evalshift-action@v0` | Runs the pipeline on pull requests, pushes the run, posts one PR comment, sets the `evalshift/regression` status. | [docs/github-action.md](docs/github-action.md) |
 | **Hosted server** — `api.evalshift.dev`, web app at `evalshift.dev` | Optional. Stores pushed run bundles, diffs them across branches, drives PR comments and gating. | [docs/hosted.md](docs/hosted.md) |
 
 The SDK and the CLI never call each other — the interface is files under
@@ -95,7 +95,7 @@ uv pip install evalshift-sdk     # or: pip install evalshift-sdk
 From source (for contributors):
 
 ```bash
-git clone https://github.com/babaliauskas/evalshift-cli.git
+git clone https://github.com/evalshift/evalshift-cli.git
 cd evalshift-cli
 uv venv --python 3.11
 source .venv/bin/activate
@@ -113,7 +113,7 @@ candidate model to it.
 evalshift init                    # minimal capture-first evalshift.yaml
 ```
 
-Instrument the agent with [evalshift-sdk](https://github.com/babaliauskas/evalshift-sdk)
+Instrument the agent with [evalshift-sdk](https://github.com/evalshift/evalshift-sdk)
 — stdlib-only, Python 3.10+, installed with the CLI or on its own:
 
 ```python
@@ -242,7 +242,7 @@ The short version:
 
 `evalshift init --ci` scaffolds a production-shaped workflow: it discovers
 every committed suite under `.evalshift/suites/`, evaluates each on every
-pull request via [`babaliauskas/evalshift-action@v0`](https://github.com/babaliauskas/evalshift-action)
+pull request via [`evalshift/evalshift-action@v0`](https://github.com/evalshift/evalshift-action)
 (one matrix job per suite), pushes the runs to hosted EvalShift, compares
 against the latest compatible base-branch run, posts one PR comment, and
 gates merges on your `migration_policy` through a single required
@@ -368,7 +368,7 @@ Runnable projects under [`examples/`](examples/):
 
 [Apache-2.0](LICENSE). Free for any use, commercial included — no share-back
 requirement, and an explicit patent grant. The capture SDK
-([`evalshift-sdk`](https://github.com/babaliauskas/evalshift-sdk)), the piece
+([`evalshift-sdk`](https://github.com/evalshift/evalshift-sdk)), the piece
 you import into your own application, is MIT.
 
 Earlier releases stay under the license they shipped with: `0.3.0` and earlier

{evalshift-1.2.0 → evalshift-1.2.2}/docs/agents.md

@@ -60,7 +60,7 @@ parsed `ToolTrace`, and the configured `tool_*` evaluators score against the
 per-example `expected_tools` ground truth.
 
 For your own agent, the recommended path is capture-first: instrument it
-with the [evalshift-sdk](https://github.com/babaliauskas/evalshift-sdk),
+with the [evalshift-sdk](https://github.com/evalshift/evalshift-sdk),
 exercise it to record captures, then `evalshift capture sync` promotes them
 into a golden suite carrying the toolset your production agent actually
 offered — see [Getting started](getting-started.md).
@@ -135,7 +135,7 @@ Nothing in `evalshift.yaml` wires a toolset to a prompt — dispatch reads it
 off each golden-suite *example* instead (`toolset_ref` or inline `tools`, see
 [Suite ground truth](#suite-ground-truth) below), so the same prompt can
 legitimately dispatch some examples with tools and others without, in one
-run. [`examples/agent/tools.yaml`](https://github.com/babaliauskas/evalshift-cli/blob/main/examples/agent/tools.yaml) is just this project's human-readable record of what
+run. [`examples/agent/tools.yaml`](https://github.com/evalshift/evalshift-cli/blob/main/examples/agent/tools.yaml) is just this project's human-readable record of what
 those tools are; it accepts either Anthropic-shape (`name` / `description` /
 `input_schema`) or OpenAI-shape (`{ "type": "function", "function": {...}
 }`) entries — `evalshift run` serialises whatever a toolset resolves to in

evalshift-1.2.2/docs/changelog.md

@@ -0,0 +1,6 @@
+# Changelog
+
+This page mirrors [`CHANGELOG.md`](https://github.com/evalshift/evalshift-cli/blob/main/CHANGELOG.md)
+in the repository.
+
+For the canonical, always-current version, see the file in the repo.

{evalshift-1.2.0 → evalshift-1.2.2}/docs/configuration.md

@@ -366,7 +366,7 @@ A list of prompt definitions. Each entry has:
 | `judge_model`   | string | `gemini-3.1-flash-lite-preview` | Default LLM-as-judge model. |
 | `insights_model`| string | (none)                        | Model that writes the run-insights narrative rendered in `report.html` and uploaded with the bundle. Falls back to `judge_model` when unset — writing analytical prose is a harder task than a pairwise A/B verdict, so it is worth tuning separately. See [Run insights](#run-insights). |
 | `concurrency`   | int    | 10 (1 ≤ x ≤ 64)               | Max in-flight LLM calls during `evalshift run` **and** `evalshift evaluate` (the embedding and judge calls made while scoring). |
-| `cache`         | bool   | `true`                        | Read/write the local SQLite cache at `~/.evalshift/cache.db`. Covers run-stage completions — for examples that offer tools, one entry per replayed round — plus `semantic` embeddings and `llm_judge` verdicts. See [Response cache](https://github.com/babaliauskas/evalshift-cli/blob/main/DOCS.md#response-cache) for the key. |
+| `cache`         | bool   | `true`                        | Read/write the local SQLite cache at `~/.evalshift/cache.db`. Covers run-stage completions — for examples that offer tools, one entry per replayed round — plus `semantic` embeddings and `llm_judge` verdicts. See [Response cache](https://github.com/evalshift/evalshift-cli/blob/main/DOCS.md#response-cache) for the key. |
 | `max_cost_usd`  | float  | 50.0                          | Soft ceiling reserved for future enforcement. The pre-flight cost prompt currently triggers above $10 (skip with `--yes`). |
 | `max_tokens`    | int    | 4096 (`> 0`)                  | Completion length cap sent to every model call. Raise it if outputs are being truncated (the provider returns `finish_reason == "length"`); a `prompts[].max_tokens` entry overrides it per prompt. Truncated calls are detected, surfaced in the report, and **excluded from the regression statistics** so a cut-off output can't manufacture a false regression. |
 | `samples_per_example` | int | 1 (1 ≤ x ≤ 20)           | How many times each `(prompt, example)` is sent to **each** model. Above 1, every sample is its own live call (the cache keys on the sample index), sample *i* of the source is scored against sample *i* of the target, and the example's row in `scores.jsonl` becomes the **mean over samples** with the per-sample scores and the within-example `delta_variance` under `metadata.samples`. The paired tests still run over examples, not samples, so this reduces noise without inflating `n`. Cost and the call count multiply by it; only worth turning on for a model that samples non-deterministically (see the report banner). See [Methodology](methodology.md#limitations-to-be-aware-of). |

{evalshift-1.2.0 → evalshift-1.2.2}/docs/getting-started.md

@@ -240,4 +240,4 @@ See [Hosted EvalShift](hosted.md) and [GitHub Action](github-action.md) for CI
 setup and privacy details.
 
 [uv]: https://docs.astral.sh/uv/
-[sdk]: https://github.com/babaliauskas/evalshift-sdk
+[sdk]: https://github.com/evalshift/evalshift-sdk

{evalshift-1.2.0 → evalshift-1.2.2}/docs/github-action.md

@@ -30,8 +30,8 @@ checklist) with three jobs:
   to the CLI that scaffolded the project: the CLI that *reads* the config in CI must be at least as new
   as the CLI that *wrote* it locally (`extra: forbid` rejects newer keys), and
   the CLI warns when the pin falls behind (or runs ahead of the local CLI) — see [Pin drift](#pin-drift). `max-parallel` defaults to 1 —
-  raise it toward your hosted plan's in-flight-run ceiling (Free 1, Pro 5,
-  Team 10). The PR comment is posted by the first matrix job only: the
+  raise it toward your hosted plan's concurrent-run limit (Pro: 5;
+  Enterprise: by contract). The PR comment is posted by the first matrix job only: the
   comment marker is a constant, so multiple suites would overwrite one
   another's summary.
 - **`evalshift gate`** — the join job to require in branch protection. It
@@ -182,12 +182,12 @@ The CLI checks for it wherever it writes or validates config — `capture sync`,
 `init` (without `--ci`, next to a workflow it did not write; `init --ci` pins
 the scaffolding CLI itself and does not warn about the file it just wrote),
 `doctor` (a `ci pin` row), and `validate`. It parses every `.github/workflows/*.yml` for
-`babaliauskas/evalshift-action` steps and compares their `evalshift-version`
+`evalshift/evalshift-action` steps and compares their `evalshift-version`
 with its own:
 
 ```text
 ⚠ CI installs evalshift 0.12.1 (.github/workflows/evalshift.yml, job evalshift) but the local CLI is 0.13.1 — an older CLI rejects config keys a newer one writes.
-  Fix: set `evalshift-version: "0.13.1"` on the babaliauskas/evalshift-action step.
+  Fix: set `evalshift-version: "0.13.1"` on the evalshift/evalshift-action step.
 ```
 
 | Status | Trigger | Fix |

{evalshift-1.2.0 → evalshift-1.2.2}/docs/hosted.md

@@ -237,20 +237,26 @@ the wrong one fails exactly like a permissions problem.
 
 Running locally is always unlimited. Plan limits apply only to what you push.
 
-When a push exceeds your organization's plan — monthly runs, seats, retention —
-or the subscription has stopped paying, the server refuses the push with HTTP
-402 and the CLI prints exactly what the server said:
+Every new hosted account gets a 30-day Pro trial (no card) that covers every
+org it creates; after that each org needs its own Pro subscription.
+
+When a push exceeds your organization's plan — monthly runs, seats, concurrent
+runs — the org's trial ended without a subscription, or the subscription has
+stopped paying, the server refuses the push with HTTP 402 and the CLI prints
+exactly what the server said:
 
 ```
 ✗ EvalShift: this run needs a paid plan.
-  Monthly run limit reached on the Free plan (50 of 50 runs used).
+  The free trial for this organization ended on 2026-11-02. Subscribe to Pro to push runs and keep the CI gate — existing runs stay readable.
   Upgrade: https://app.evalshift.dev/app/acme/settings/billing
 ```
 
 `push` exits 1 and nothing is uploaded. The CLI does not evaluate entitlements
 itself and does not retry a payment error — retrying cannot change the answer.
 Your run and its `report.html` stay on disk under `.evalshift/runs/`, so you can
-push the same run id again after upgrading, or after the monthly counter resets.
+push the same run id again once the org is subscribed (or, for the monthly
+quota, after the counter resets). An expired org stays readable — its runs,
+diffs and baselines still load.
 
 Transient upload failures (HTTP 429 and 5xx from object storage) are a separate
 case and *are* retried with exponential backoff.
@@ -355,5 +361,5 @@ credential file locally and repository secrets in CI.
 | `could not determine a valid 40-character git SHA` | `bundle` (or `push <run-id>` building a bundle) ran outside a git checkout, or in one with no commits, and `GITHUB_SHA` is unset. | Run inside a git repository with at least one commit, or set `GITHUB_SHA`. |
 | `project was not found` | The project does not exist and auto-create is disabled or not allowed. | Ask an owner to create it, use an org-scoped owner token, or enable auto-create. |
 | `cannot auto-create <slug> at <host>` | The message names the host it talked to and the server's status. Most often the host is not the one you meant: with no `--host` and no `EVALSHIFT_HOST`, an unset credentials file falls back to `https://api.evalshift.dev`, where your org does not exist. | Run `evalshift whoami` and check the host it prints. If it is wrong, `evalshift login --host <hosted-api-url>`. If the host is right and the status is 403, the token lacks org access — see [Project auto-create](#project-auto-create). |
-| `this run needs a paid plan` | The org's plan does not cover this push, or the subscription has stopped paying. | Open the upgrade URL printed with the message, or wait for the monthly reset and push the same run id again. See [Plan limits](#plan-limits). |
+| `this run needs a paid plan` | The org's plan does not cover this push, its 30-day trial ended without a subscription, or the subscription has stopped paying. | Open the upgrade URL printed with the message; for the monthly quota you can also wait for the reset and push the same run id again. See [Plan limits](#plan-limits). |
 | `this run carries no migration policy` warning | No `migration_policy` is configured in `evalshift.yaml`, so the bundle has no `decision.policy`. | Add `migration_policy` to `evalshift.yaml` (see [Configuration](configuration.md#migration_policy)). Until then the gate has only whatever old web-app policy the project still has; with none, it reports `inconclusive` and never blocks the pull request unless the GitHub Action sets `require-policy: true`. |

{evalshift-1.2.0 → evalshift-1.2.2}/docs/sdk.md

@@ -1,13 +1,13 @@
 # Capture SDK
 
-The [evalshift-sdk](https://github.com/babaliauskas/evalshift-sdk) is a separate
+The [evalshift-sdk](https://github.com/evalshift/evalshift-sdk) is a separate
 package (`pip install evalshift-sdk`, import name `evalshift`) that you install
 **inside your agent process**. It records what your agent actually did — model
 calls, tool calls, the final output — as JSON capture files on disk. The CLI
 promotes those captures into golden suites.
 
 This page covers the CLI side of that contract. The full SDK guide lives in the
-SDK repo: [DOCS.md](https://github.com/babaliauskas/evalshift-sdk/blob/main/DOCS.md),
+SDK repo: [DOCS.md](https://github.com/evalshift/evalshift-sdk/blob/main/DOCS.md),
 dense LLM reference at <https://www.evalshift.dev/sdk-llms-full.txt>.
 
 ## The contract
@@ -61,7 +61,7 @@ to state its masking policy in the call itself. `True` applies the SDK's
 `default_redactor` (emails, `sk-…`, `AKIA…`, `Bearer …`), `False` records
 verbatim, and a `(value) -> value` callable does something custom; any other
 value — `None` included — raises `TypeError`. Details:
-[REDACTION.md](https://github.com/babaliauskas/evalshift-sdk/blob/main/docs/REDACTION.md).
+[REDACTION.md](https://github.com/evalshift/evalshift-sdk/blob/main/docs/REDACTION.md).
 
 `tools` is required on the same entry points: the toolset the agent was offered,
 or `[]` if it never calls tools. Omitting it is a `TypeError` too.
@@ -131,7 +131,7 @@ carries the recorded tool results so `run` replays later rounds teacher-forced
 `--keep-duplicates` disables dedup.
 
 Full behaviour: [Configuration](configuration.md) and the
-[Capturing from production](https://github.com/babaliauskas/evalshift-cli/blob/main/DOCS.md#capturing-from-production)
+[Capturing from production](https://github.com/evalshift/evalshift-cli/blob/main/DOCS.md#capturing-from-production)
 section of DOCS.md.
 
 ## 4. Evaluate a candidate against real behaviour
@@ -145,7 +145,7 @@ from production traffic rather than hand-written examples.
 
 ## Related
 
-- [`examples/capture-first/`](https://github.com/babaliauskas/evalshift-cli/tree/main/examples/capture-first)
+- [`examples/capture-first/`](https://github.com/evalshift/evalshift-cli/tree/main/examples/capture-first)
   — every step above checked in: the instrumented agent, the captures and
   toolset sidecar it wrote, the promoted suite, and the `evalshift.yaml` whose
   managed `suites:` block `capture sync` filled in.

{evalshift-1.2.0 → evalshift-1.2.2}/examples/capture-first/README.md

@@ -1,7 +1,7 @@
 # Capture-first example
 
 The flow the README and `evalshift init` recommend, checked in end to end:
-instrument an agent with the [evalshift-sdk](https://github.com/babaliauskas/evalshift-sdk),
+instrument an agent with the [evalshift-sdk](https://github.com/evalshift/evalshift-sdk),
 exercise it, then let `evalshift capture sync` turn what it recorded into a
 golden suite and wire that suite into `evalshift.yaml`.
 

{evalshift-1.2.0 → evalshift-1.2.2}/examples/capture-first/evalshift.yaml

@@ -1,5 +1,5 @@
 # migration_profile: model-upgrade
-# EvalShift configuration. See https://github.com/babaliauskas/EvalShift for docs.
+# EvalShift configuration. See https://github.com/evalshift/evalshift-cli for docs.
 #
 # `init` writes only this file, set up for the capture-first flow: instrument
 # your agent with the evalshift-sdk, exercise it to record captures, then run

{evalshift-1.2.0 → evalshift-1.2.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "evalshift"
-version = "1.2.0"
+version = "1.2.2"
 description = "Run your prompts on two LLMs and find out, with statistical confidence, what regressed."
 readme = "README.md"
 license = "Apache-2.0"
@@ -30,7 +30,7 @@ dependencies = [
     # The capture SDK (import name `evalshift`). Stdlib-only, so it adds nothing to
     # the environment; declared so one `pip install evalshift` serves both
     # instrumenting an agent and running evaluations, and so `import evalshift` is
-    # always the SDK. See docs/superpowers/specs/2026-09-09-namespace-collision-design.md.
+    # always the SDK. See the co-install note in DOCS.md.
     "evalshift-sdk>=0.4.0",
     "typer>=0.12",
     "pydantic>=2.5",
@@ -68,11 +68,11 @@ dev = [
 ]
 
 [project.urls]
-Homepage = "https://github.com/babaliauskas/evalshift-cli"
-Documentation = "https://github.com/babaliauskas/evalshift-cli#readme"
-Repository = "https://github.com/babaliauskas/evalshift-cli"
-Issues = "https://github.com/babaliauskas/evalshift-cli/issues"
-Changelog = "https://github.com/babaliauskas/evalshift-cli/blob/main/CHANGELOG.md"
+Homepage = "https://github.com/evalshift/evalshift-cli"
+Documentation = "https://github.com/evalshift/evalshift-cli#readme"
+Repository = "https://github.com/evalshift/evalshift-cli"
+Issues = "https://github.com/evalshift/evalshift-cli/issues"
+Changelog = "https://github.com/evalshift/evalshift-cli/blob/main/CHANGELOG.md"
 
 [project.scripts]
 evalshift = "evalshift_cli.cli.main:app"

{evalshift-1.2.0 → evalshift-1.2.2}/src/evalshift_cli/cli/commands/_scaffold.py

@@ -231,8 +231,8 @@ jobs:
     strategy:
       # One suite regressing must not cancel the others' results.
       fail-fast: false
-      # Hosted EvalShift refuses uploads past your org plan's in-flight run
-      # ceiling (Free 1, Pro 5, Team 10). Raise this toward your plan's limit
+      # Hosted EvalShift refuses uploads past your org's concurrent-run
+      # limit (Pro: 5; Enterprise: by contract). Raise this toward it
       # to evaluate suites in parallel.
       max-parallel: 1
       matrix:
@@ -251,7 +251,7 @@ jobs:
 
       - name: Run evalshift on ${{ matrix.suite }}
         if: ${{ env.HAS_EVALSHIFT_TOKEN == 'true' }}
-        uses: babaliauskas/evalshift-action@v0
+        uses: evalshift/evalshift-action@v0
         with:
           token: ${{ secrets.EVALSHIFT_TOKEN }}
           config: evalshift.yaml

{evalshift-1.2.0 → evalshift-1.2.2}/src/evalshift_cli/cli/commands/init.py

@@ -98,7 +98,7 @@ _SEMANTIC_BLOCK_DISABLED: Final = """\
 # Body of the minimal config, up to (but excluding) the suites region and the
 # migration_policy block. ``render_minimal_config`` appends those.
 _MINIMAL_YAML_BODY: Final = """\
-# EvalShift configuration. See https://github.com/babaliauskas/EvalShift for docs.
+# EvalShift configuration. See https://github.com/evalshift/evalshift-cli for docs.
 #
 # `init` writes only this file, set up for the capture-first flow: instrument
 # your agent with the evalshift-sdk, exercise it to record captures, then run

{evalshift-1.2.0 → evalshift-1.2.2}/src/evalshift_cli/hosted/bundle_manifest.schema.json

@@ -1018,7 +1018,7 @@
       },
       "GuardrailEvent": {
         "additionalProperties": false,
-        "description": "A guardrail or policy check. Imported agent traces only.",
+        "description": "A guardrail or policy check.\n\nReserved for a bring-your-own-agent trace; no current producer emits it \u2014\n``evalshift traces import`` writes such traces only to a local artifact,\nnot into a pushed bundle.",
         "properties": {
           "name": {
             "minLength": 1,
@@ -1405,7 +1405,7 @@
       },
       "RetrievalEvent": {
         "additionalProperties": false,
-        "description": "A retrieval step. Imported agent traces only.",
+        "description": "A retrieval step.\n\nReserved for a bring-your-own-agent trace; no current producer emits it \u2014\n``evalshift traces import`` writes such traces only to a local artifact,\nnot into a pushed bundle.",
         "properties": {
           "documents": {
             "items": {

{evalshift-1.2.0 → evalshift-1.2.2}/src/evalshift_cli/hosted/push.py

@@ -108,6 +108,39 @@ def _upgrade_prompt(exc: HostedHTTPError) -> PushError:
     return PushError("\n".join(lines))
 
 
+def _initiate_after_create(
+    client: HostedClient, manifest: dict[str, Any], *, size_bytes: int
+) -> dict[str, Any]:
+    """The ``POST /runs`` retried after auto-creating the project, mapped like the first.
+
+    It runs inside the first attempt's ``except`` clause, so its own errors would skip the
+    sibling handlers and surface as a traceback. ``POST /runs`` answers 404 for a missing
+    project before any entitlement check, so a 402 for the monthly run or concurrent-run
+    limit, seat overage, or a subscription that stopped paying arrives on this retry, not
+    on the first call.
+
+    Args:
+        client: The hosted client the first attempt used.
+        manifest: The bundle manifest sent as the run's create payload.
+        size_bytes: Size of the bundle file that will be uploaded.
+
+    Returns:
+        The server's ``POST /runs`` response.
+
+    Raises:
+        PushError: On any hosted error — the upgrade prompt for a 402, the server's
+            message otherwise.
+    """
+    try:
+        return client.initiate_run(manifest, size_bytes=size_bytes)
+    except HostedHTTPError as exc:
+        if exc.status_code == _PAYMENT_REQUIRED:
+            raise _upgrade_prompt(exc) from exc
+        raise PushError(str(exc)) from exc
+    except (HostedNetworkError, HostedError) as exc:
+        raise PushError(str(exc)) from exc
+
+
 @dataclass(frozen=True, slots=True)
 class PushResult:
     """Outcome of a hosted push.
@@ -239,7 +272,7 @@ def push_bundle(
             raise PushError(str(exc)) from exc
         _auto_create_project(client, project_slug=str(manifest["project_slug"]))
         project_created = True
-        response = client.initiate_run(manifest, size_bytes=upload_size_bytes)
+        response = _initiate_after_create(client, manifest, size_bytes=upload_size_bytes)
     except (HostedNetworkError, HostedError) as exc:
         raise PushError(str(exc)) from exc
 
@@ -489,17 +522,25 @@ def _auto_create_project(client: HostedClient, *, project_slug: str) -> None:
             f"permission — check `evalshift whoami` reports this host and an org "
             f"named {org_slug!r}",
         ) from exc
+    except (HostedNetworkError, HostedError) as exc:
+        raise PushError(str(exc)) from exc
     if any(item.get("slug") == project for item in projects):
         return
     try:
         client.create_project(org_slug, slug=project, name=_name_from_slug(project))
     except HostedHTTPError as exc:
+        if exc.status_code == _PAYMENT_REQUIRED:
+            # An expired org is read-only: creating a project is refused with the same 402 a
+            # push gets, and it deserves the same prompt rather than an access-permission hint.
+            raise _upgrade_prompt(exc) from exc
         raise PushError(
             f"cannot auto-create {project_slug!r} at {client.host}: "
             f"{_server_said(exc)}. Creating a project needs owner access to "
             f"{org_slug!r}; a scoped service-account key cannot do it — create the "
             f"project by hand, or push with an owner token",
         ) from exc
+    except (HostedNetworkError, HostedError) as exc:
+        raise PushError(str(exc)) from exc
 
 
 def _server_said(exc: HostedHTTPError) -> str:

{evalshift-1.2.0 → evalshift-1.2.2}/src/evalshift_cli/hosted/trace_events.py

@@ -12,6 +12,11 @@ Two internal representations describe that, and neither goes on the wire:
 
 Both map here onto one wire shape so the server, the database and the client
 speak a single language. Neither internal model changes.
+
+Only :func:`from_tool_trace` has a production caller (``hosted/bundle.py``): a run
+bundle's ``examples[].traces`` come from the replay's own tool-call traces.
+:func:`from_agent_trace` converts imported agent traces to the same wire shape but is
+not wired into bundling today — those traces stay local.
 """
 
 from __future__ import annotations
@@ -170,7 +175,14 @@ _PASSTHROUGH_FIELDS: dict[str, tuple[str, ...]] = {
 
 
 def from_agent_trace(trace: AgentTrace) -> dict[str, Any] | None:
-    """Wire stream for one imported bring-your-own-agent trace.
+    """Convert an imported agent trace to bundle wire events.
+
+    Not called by ``bundle.py`` today: imported traces (``evalshift traces import``)
+    stay local in ``.evalshift/runs/<id>/traces.jsonl`` and are read only by
+    ``evaluate``, ``report.json`` and the ``diff``/``inspect``/``replay case``
+    commands. A run bundle's ``examples[].traces`` come solely from the replay's own
+    tool-call traces via :func:`from_tool_trace`. This function exists for a future
+    caller that would include imported traces in the bundle.
 
     A round begins at each ``model_call``. The counter increments on every
     ``model_call`` after the first, so events preceding any model call stay in

{evalshift-1.2.0 → evalshift-1.2.2}/src/evalshift_cli/reports/templates/report.css

@@ -556,6 +556,33 @@ details[open] > summary .caret::after { content: "\25BE"; }
   background: var(--strip);
   flex-wrap: wrap;
 }
+/* The row must read as a control, not a static line: a boxed caret, a worded
+   hint, and a hover state that reacts across the whole header. */
+.reg-head .caret {
+  display: inline-flex;
+  align-items: center;
+  justify-content: center;
+  width: 22px;
+  height: 22px;
+  flex: none;
+  border: 1px solid var(--stroke);
+  background: var(--panel-2);
+}
+.reg-head .caret::after { font-size: 14px; width: auto; color: var(--body); }
+.reg-toggle {
+  font-family: var(--font-mono);
+  font-size: 12px;
+  color: var(--dim);
+  min-width: 88px;
+  text-align: right;
+}
+.reg-toggle::after { content: "Show details"; }
+.regression[open] > .reg-head .reg-toggle::after { content: "Hide details"; }
+.reg-head:hover { background: var(--panel-2); }
+.reg-head:hover .caret { border-color: var(--ok-edge); }
+.reg-head:hover .caret::after,
+.reg-head:hover .reg-toggle { color: var(--accent); }
+.reg-head:focus-visible { outline: 2px solid var(--accent); outline-offset: -2px; }
 .reg-id { font-family: var(--font-mono); font-size: 14px; color: var(--fg); }
 .reg-what {
   font-size: 14px;
@@ -762,6 +789,8 @@ footer p { margin: 0; }
      id instead. */
   .reg-head { gap: 8px 12px; padding: 12px 14px; }
   .reg-what { flex: 1 1 60%; min-width: 0; white-space: normal; }
+  /* The boxed caret alone carries the affordance when space is tight. */
+  .reg-toggle { display: none; }
   .reg-body { padding: 14px; }
   .trace-list { padding: 12px 12px 12px 26px; }
   .transcript-body { padding: 0 12px 12px 28px; }
@@ -769,6 +798,10 @@ footer p { margin: 0; }
 
 @media print {
   body { background: #fff; color: #111; }
+  /* Paper can't be clicked: show every regression body. Best-effort: only
+     engines with ::details-content (Chromium 131+) honour it. */
+  .regression::details-content { content-visibility: visible; }
+  .reg-toggle { display: none; }
 }
 
 /* --- Tone utilities ------------------------------------------------------ */