auditkit 1.1.0__tar.gz → 1.1.2__tar.gz

{auditkit-1.1.0 → auditkit-1.1.2}/CHANGELOG.md

@@ -11,19 +11,18 @@ All notable changes to AuditKIT. The format follows [Keep a Changelog](https://k
   SGLang), e.g. `{"enable_thinking": False}` for Qwen3. A request's own values win.
   Part of the run fingerprint.
 - `api:` says when a server may have ignored `chat_template_kwargs` or
-  `parallel_tool_calls` (#43): `RunResult.metadata["unverified_request_fields"]`, a
+  `parallel_tool_calls`: `RunResult.metadata["unverified_request_fields"]`, a
   `summary()` line, one warning, and `cap_unverified` on the parallel scores. The
   server is probed once (`GET /models` `owned_by`) and recorded in
   `metadata["api_server"]`; only `APIModel(server="sglang"|"vllm")` clears the note.
   New `Model.run_notes()` hook for backend facts about a run.
-- `ak.load_bfcl()` and `ak.bfcl_arg_match` (#45 phase 2): BFCL v3's 11 single-turn
+- `ak.load_bfcl()` and `ak.bfcl_arg_match`: BFCL v3's 11 single-turn
   categories plus the two irrelevance ones, read from the Hub files (JSON Lines). The
   accepted-value lists stay on the reference, and `bfcl_arg_match` compares against
   them as BFCL's AST checker does. It handles a drifted answer id, flags unsatisfiable
   answers, passes system prompts through, converts BFCL types to JSON Schema, and has an
   optional `dots_to_underscores`. Multi-turn is refused: its answers are call strings.
-  Counts at the pinned revision are in `docs/notes/bfcl-possible-answer-counts.md`.
-- `RunResult.route_distribution()` and a `summary()` line (#45): for samples whose
+- `RunResult.route_distribution()` and a `summary()` line: for samples whose
   reference names several routes (`any_of`), which route each took and which rule
   decided (`f1`, `recall`, `turn_structure`, `call_order`, `listed_order`), with the
   ties broken by listed order counted. Information only: it never reaches the stats.
@@ -41,7 +40,7 @@ All notable changes to AuditKIT. The format follows [Keep a Changelog](https://k
   layers' own `config.sliding_window`. Rule (m) does the same for SGLang's generic
   Transformers backend (Aya Vision: Triton decode crashed with `kv_indptr=None`),
   returning a window only for models with sliding-window layers.
-- The `vllm:` column covers all nine Cohere models (#52): the full matrix ran on an RTX PRO 6000
+- The `vllm:` column covers all nine Cohere models: the full matrix ran on an RTX PRO 6000
   (CUDA 12.8) with the FlashInfer sampler fix, every check passing. Aya Expanse 32B and Aya
   Vision 32B score 5/5; the other seven match their A100 scores exactly.
 - `vllm:` on SM 12.x GPUs (RTX PRO 6000, RTX 50-series) with a CUDA toolkit older than 12.9:
@@ -61,18 +60,18 @@ All notable changes to AuditKIT. The format follows [Keep a Changelog](https://k
   32B: all nine Cohere models now run on SGLang 0.5.20. `colab_sglang_32b_north.ipynb`
   measures the two.
 - `docs/model_backends.md`: the `vllm:` column is measured for the seven Cohere models that fit
-  a 40 GB A100 (#52; vLLM 0.30.0): Tiny Aya 4/5 each, North 5/5, Aya Expanse 8B and Aya Vision
+  a 40 GB A100 (vLLM 0.30.0): Tiny Aya 4/5 each, North 5/5, Aya Expanse 8B and Aya Vision
   8B 5/5, image checks passing. vLLM falls back to its Transformers backend for Aya Vision on
   its own.
   "Aya Vision loads on neither vLLM nor SGLang" is corrected: it loads on both, non-natively.
-  The images section and README now say `api:` sends images (since #50) and the offline
-  `vllm:` backend still drops them; the install trap names #51's cause (a stale auditkit 1.0.0).
+  The images section and README now say `api:` sends images and the offline
+  `vllm:` backend still drops them; the install trap names the cause (a stale auditkit 1.0.0).
 - `check_compat()` rule (j): with transformers >= 5.15 in the SGLang env (needed for North
   Micro Vision), SGLang's Transformers backend rejects the `embedding_rowwise` TP style
   that transformers adds for tied embeddings, so North still fails to load. The entry is
   inert there (only `nn.Linear` reads the plan); `--patch-sglang` maps it to `replicate`.
 - transformers' "does not recognize this architecture" is rewritten by `hf:` and `vllm:` to
-  name the model type, the installed transformers and the version it needs (#51):
+  name the model type, the installed transformers and the version it needs:
   "'…North-Micro-Vision-Instruct' is a 'cohere_compass' model, and the installed
   transformers 4.57.6 does not know that architecture: it needs transformers>=5.15. The
   checkpoint is fine." The original read like a broken checkpoint. The real cause was a
@@ -108,15 +107,27 @@ All notable changes to AuditKIT. The format follows [Keep a Changelog](https://k
 
 ### Fixed
 
+- `auditkit eval --config FILE` accepts `model:` from the config file. `--model` was
+  `required=True` on the parser, so argparse rejected the command with "the following
+  arguments are required: --model" before the YAML was read, even though `docs/cli.md`
+  documents exactly that invocation. It is now optional on the parser and checked after the
+  config merge; a genuinely missing model still exits 2, naming both ways to supply it.
 - The six built-in benchmark scenarios load again. They used bare Hub ids (`mmlu`, `arc` and
   `truthfulqa` no longer exist; the others failed with `HfUriError`) and now use `cais/mmlu`,
   `openai/gsm8k`, `allenai/ai2_arc`, `Rowan/hellaswag`, `truthfulqa/truthful_qa` and
   `openai/openai_humaneval`. HumanEval also referenced a `TaskKind.CODE` that doesn't exist; it loads
   as a generative task (no code-execution metric, so use `ak.run_lmeval("humaneval", ...)` for pass@k).
+- Docs: the scorer reference and metrics overview cover all 68 registered metrics, and the
+  data model documents `Sample.images`, `RunConfig.seed`/`track_performance`/
+  `chat_template_kwargs` and `RunResult.unscored`/`metadata`; every `eval` flag is in
+  `docs/cli.md`; the extras table matches `pyproject.toml`; the claimed-but-unimplemented
+  features are gone, as are the internal pages that were still being published to the site.
+- The `dev` extra declares `pyyaml`, which the `--config` tests need. PyYAML remains an
+  optional runtime dependency; the core install is still dependency-free.
 - Docs: no METEOR (never implemented) and no "Rich HTML reports" tick (not implemented); current
   Claude model ids instead of retired Claude 3 ones; "source-available", not "open-source"; no
-  "hackathon" wording, links to a private repository or personal-repo issue numbers in public pages;
-  GUIDE's table of contents, catalog heading and version line match its sections and 1.2.0.
+  internal wording, links to a private repository or private issue numbers in public pages;
+  GUIDE's table of contents, catalog heading and version line match its sections and 1.1.2.
 
 ### Changed
 
@@ -125,29 +136,28 @@ All notable changes to AuditKIT. The format follows [Keep a Changelog](https://k
   "triangle". **Scores change.** `F1Score(normalize=False)` keeps the old raw
   whitespace split; the setting is part of the run fingerprint.
 - `[lmeval]` extra: `accelerate>=1.0`.
-- **`abstention`: `short_reply_markers` now takes regular expressions** (#35). The
+- **`abstention`: `short_reply_markers` now takes regular expressions**. The
   defaults require refusal-shaped wording ("insufficient information to/for/in/about…",
   "no relevant information/documents/context…"), so a short real answer such as
   "Yes, there is no relevant fee for domestic wires." is no longer a refusal, and a
   first sentence opening with "yes" never is. Bare phrases passed to
   `Abstention(short_reply_markers=...)` still work (a plain phrase is a valid regex);
   a phrase containing regex metacharacters must be escaped (`re.escape`).
-- **An unnamed recorded tool call is unscored, not a measured 0.0** (#36). An EventLog
+- **An unnamed recorded tool call is unscored, not a measured 0.0**. An EventLog
   or AgentTune trajectory action that carries call arguments (`arguments`,
   `parameters`, `args`, `input`, `query`) but no tool name leaves
   `tool_call_names` unavailable, so `tool_call_f1` / `trajectory_match` /
   `parallel_tool_calls` are skipped, and `rescore` reports `trace.tool_call_names`.
-- **A trajectory whose actions are all empty is a measured zero calls** (#37): it
+- **A trajectory whose actions are all empty is a measured zero calls**: it
   scores (0.0 against expected calls, 1.0 on a no-call case) instead of being dropped.
 - Docs: SGLang needs Python 3.10-3.12 (a uv 3.12 venv on 3.13 hosts); example 13
   builds its SGLang venv that way. `docs/VLLM_KNOWN_ISSUES.md` covers the Colab
   torchaudio mismatch after `[vllm]`.
 
-## [1.2.0] - 2026-09-28
+## [1.1.2] - 2026-10-02
 
-Lexsi stack interop for the hackathon tracks (CuratorKIT, AlignTune, SafeTune,
-CircuitKIT and AgentTune outputs evaluate with no glue code), plus the
-hackathon readiness fixes for the permitted Cohere models.
+Lexsi stack interop (CuratorKIT, AlignTune, SafeTune, CircuitKIT and AgentTune
+outputs evaluate with no glue code), plus fixes for the Cohere models.
 
 ### Added
 
@@ -166,7 +176,7 @@ hackathon readiness fixes for the permitted Cohere models.
 - **Native tool calling on `hf:`.** Tool schemas go to the chat template
   (`apply_chat_template(tools=...)`), so `ToolCallAdapter()` works on local
   models whose template renders tools; a template without tool support raises
-  `CapabilityError`. None of the Cohere hackathon models' templates (Tiny Aya,
+  `CapabilityError`. None of the Cohere models' templates (Tiny Aya,
   Aya Expanse, Aya Vision, North) render tools: use
   `ToolCallAdapter(mode="prompt")` with them.
 - **Cohere tool-call format.** The tool-call parser reads Command R7B
@@ -189,11 +199,11 @@ hackathon readiness fixes for the permitted Cohere models.
 ### Fixed
 
 - A JSON-string reference (a CSV cell, a JSONL field) now fingerprints like its
-  decoded form (#44), in the dataset fingerprint and in `AgentCase.digest()`, so the
+  decoded form, in the dataset fingerprint and in `AgentCase.digest()`, so the
   same dataset read from a file reuses the cached run. Dict and list references keep
   their digests. `auditkit agent --dry-run` shows `refs=yes` for an explicit `[]`
   reference ("expect no call").
-- A bare prefix is now decided per backend (#43). `api:`, `hf:` and `vllm:` raise
+- A bare prefix is now decided per backend. `api:`, `hf:` and `vllm:` raise
   "empty model spec" (`api:` used to send `{"model": null}`). The hosted prefixes
   (`openai:`, `anthropic:`, `groq:`, `openrouter:`, `lexsi:`, `litellm:`) fall back
   to their backend's default model, and `agent:` still takes its URL from `url=`.

{auditkit-1.1.0 → auditkit-1.1.2}/CITATION.cff

@@ -8,7 +8,7 @@ abstract: |
 type: software
 message: "If you use this software, please cite it."
 
-version: 1.1.0
+version: 1.1.2
 
 authors:
   - family-names: "Arora"

{auditkit-1.1.0 → auditkit-1.1.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: auditkit
-Version: 1.1.0
+Version: 1.1.2
 Summary: Evaluate any model on any dataset and any task — standalone, no platform required
 Project-URL: Homepage, https://github.com/Lexsi-Labs/AuditKIT
 Project-URL: Repository, https://github.com/Lexsi-Labs/AuditKIT
@@ -126,6 +126,7 @@ Requires-Dist: pillow; extra == 'all'
 Requires-Dist: pytest; extra == 'all'
 Requires-Dist: pytest-asyncio; extra == 'all'
 Requires-Dist: pytest-mock; extra == 'all'
+Requires-Dist: pyyaml; extra == 'all'
 Requires-Dist: requests; extra == 'all'
 Requires-Dist: torch>=2.4; extra == 'all'
 Requires-Dist: torchmetrics; extra == 'all'
@@ -142,6 +143,7 @@ Provides-Extra: dev
 Requires-Dist: pytest; extra == 'dev'
 Requires-Dist: pytest-asyncio; extra == 'dev'
 Requires-Dist: pytest-mock; extra == 'dev'
+Requires-Dist: pyyaml; extra == 'dev'
 Provides-Extra: interop
 Requires-Dist: datasets; extra == 'interop'
 Requires-Dist: gitpython; extra == 'interop'
@@ -180,8 +182,8 @@ Description-Content-Type: text/markdown
 <p align="center">
   <a href="https://github.com/Lexsi-Labs/AuditKIT">
     <picture>
-      <source media="(prefers-color-scheme: dark)" srcset="docs/assets/auditkit-logo-white.png">
-      <img src="docs/assets/auditkit-logo-black.png" alt="AuditKit" width="480">
+      <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/Lexsi-Labs/AuditKIT/main/docs/assets/auditkit-logo-white.png">
+      <img src="https://raw.githubusercontent.com/Lexsi-Labs/AuditKIT/main/docs/assets/auditkit-logo-black.png" alt="AuditKit" width="480">
     </picture>
   </a>
 </p>
@@ -193,7 +195,7 @@ Description-Content-Type: text/markdown
 </p>
 
 <p align="center">
-  <a href="https://pypi.org/project/auditkit/"><img src="https://img.shields.io/badge/pypi-v1.2.0-0a8868" alt="PyPI v1.2.0"></a>
+  <a href="https://pypi.org/project/auditkit/"><img src="https://img.shields.io/pypi/v/auditkit?color=0a8868" alt="PyPI version"></a>
   <a href="https://www.python.org/"><img src="https://img.shields.io/badge/python-3.10%2B-blue" alt="Python 3.10+"></a>
   <a href="https://github.com/Lexsi-Labs/AuditKIT/blob/main/LICENSE.md"><img src="https://img.shields.io/badge/license-LSAL--1.2-lightgrey" alt="License: LSAL-1.2 (source-available, noncommercial)"></a>
   <a href="https://auditkit.lexsi.ai/"><img src="https://img.shields.io/badge/docs-auditkit.lexsi.ai-4c6ef5" alt="Documentation"></a>
@@ -219,7 +221,7 @@ Description-Content-Type: text/markdown
 | **Models** | 10 backends: OpenAI, Anthropic, HuggingFace, Lexsi, vLLM, LiteLLM, API, Groq, OpenRouter, Agent (HTTP endpoint) — all resolved via `AutoModel.resolve()`. Any `list[str] → list[str]` callable also works. |
 | **Zero deps** | Core runs on stdlib. Backends and heavy metrics are optional extras (`pip install auditkit[openai]`). |
 | **Fingerprints** | Every run gets a stable sha256 — results are cacheable, comparable, and reproducible by construction. |
-| **Status** | ~3,000 tests, v1.2.0, LSAL-1.2 license (source-available, noncommercial). 11 metric families, 6 CLI subcommands, YAML config, MKDocs site. |
+| **Status** | ~3,000 tests, v1.1.2, LSAL-1.2 license (source-available, noncommercial). 11 metric families, 6 CLI subcommands, YAML config, MKDocs site. |
 
 ---
 
@@ -359,7 +361,7 @@ dropped it after 0.24.0, SGLang never had it), so both servers run it through th
 Transformers backends: automatically on vLLM, with `compat.py --patch-sglang` on SGLang. North Micro Vision is native on vLLM; on SGLang
 it needs transformers >= 5.15 in the SGLang env plus `--patch-sglang`. Tiny Aya and Aya Expanse
 are native on both. The full
-model-by-runtime table is in [docs/model_backends.md](docs/model_backends.md).
+model-by-runtime table is in [docs/model_backends.md](https://github.com/Lexsi-Labs/AuditKIT/blob/main/docs/model_backends.md).
 
 Tiny Aya (`CohereLabs/tiny-aya-global` and siblings `fire`, `water`, `earth`),
 Aya Expanse and Aya Vision are gated
@@ -392,57 +394,57 @@ Every `evaluate()` returns a `RunResult`:
 ## Examples
 
 Colab-ready notebooks with real models and real datasets — see
-[`examples/README.md`](examples/README.md) for the full, current list.
+[`examples/README.md`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/README.md) for the full, current list.
 
 | Example | File |
 |---------|------|
-| Full pipeline: adapter, 5 metrics, LLM judge, LLM annotator | [`01_full_evaluation_pipeline.ipynb`](examples/01_full_evaluation_pipeline.ipynb) |
-| Generation across two real HF model families, compared | [`02_generation_across_hf_families.ipynb`](examples/02_generation_across_hf_families.ipynb) |
-| Custom annotators (regex, LLM-backed, fully custom) | [`03_custom_annotators.ipynb`](examples/03_custom_annotators.ipynb) |
-| Metrics deep dive: built-in, custom, LLM-as-judge, RAG | [`04_metrics_deep_dive.ipynb`](examples/04_metrics_deep_dive.ipynb) |
-| Every data type/task kind: generative, MCQ, RAG, precomputed, chat | [`05_data_types.ipynb`](examples/05_data_types.ipynb) |
-| Model comparison deep dive: `compare_models()` + `RunComparison` | [`06_model_comparison.ipynb`](examples/06_model_comparison.ipynb) |
-| Annotators across 4 real model families, then compared together | [`07_annotators_across_models.ipynb`](examples/07_annotators_across_models.ipynb) |
-| `CompareResult` deep dive: what `compare_models()`'s native per-model support still can't express (different scorers per model), full method surface | [`08_compare_result_deep_dive.ipynb`](examples/08_compare_result_deep_dive.ipynb) |
-| `GuardJudge` implementation check: all 5 profiles, real and offline | [`09_guard_judge_implementation_check.ipynb`](examples/09_guard_judge_implementation_check.ipynb) |
-| `GuardJudge` with real, flagship guard models | [`10_guard_judge_legit_models.ipynb`](examples/10_guard_judge_legit_models.ipynb) |
-| `EncoderJudge`: the base mechanism plus its two prebuilt subclasses | [`11_encoder_judge_prebuilts.ipynb`](examples/11_encoder_judge_prebuilts.ipynb) |
-| Performance metrics on a real, large-scale evaluation | [`12_performance_metrics_demo.ipynb`](examples/12_performance_metrics_demo.ipynb) |
-| SGLang: environment checks, server mode via `api:`, tool calls and parallel tool calls | [`13_sglang_compatibility.ipynb`](examples/13_sglang_compatibility.ipynb) |
-| Agent and RAG evals: tool calls, parallel tool calls, deployed agent endpoints, AgentTune data, retrieval and judged RAG metrics | [`14_agent_and_rag_evals.ipynb`](examples/14_agent_and_rag_evals.ipynb) |
-| End-to-end agent evaluation, fully offline (`.py` script): recorded episodes, a verified state outcome, and a fabricated answer that does not pass a state case | [`agent_eval_offline.py`](examples/agent_eval_offline.py) |
+| Full pipeline: adapter, 5 metrics, LLM judge, LLM annotator | [`01_full_evaluation_pipeline.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/01_full_evaluation_pipeline.ipynb) |
+| Generation across two real HF model families, compared | [`02_generation_across_hf_families.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/02_generation_across_hf_families.ipynb) |
+| Custom annotators (regex, LLM-backed, fully custom) | [`03_custom_annotators.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/03_custom_annotators.ipynb) |
+| Metrics deep dive: built-in, custom, LLM-as-judge, RAG | [`04_metrics_deep_dive.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/04_metrics_deep_dive.ipynb) |
+| Every data type/task kind: generative, MCQ, RAG, precomputed, chat | [`05_data_types.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/05_data_types.ipynb) |
+| Model comparison deep dive: `compare_models()` + `RunComparison` | [`06_model_comparison.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/06_model_comparison.ipynb) |
+| Annotators across 4 real model families, then compared together | [`07_annotators_across_models.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/07_annotators_across_models.ipynb) |
+| `CompareResult` deep dive: what `compare_models()`'s native per-model support still can't express (different scorers per model), full method surface | [`08_compare_result_deep_dive.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/08_compare_result_deep_dive.ipynb) |
+| `GuardJudge` implementation check: all 5 profiles, real and offline | [`09_guard_judge_implementation_check.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/09_guard_judge_implementation_check.ipynb) |
+| `GuardJudge` with real, flagship guard models | [`10_guard_judge_legit_models.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/10_guard_judge_legit_models.ipynb) |
+| `EncoderJudge`: the base mechanism plus its two prebuilt subclasses | [`11_encoder_judge_prebuilts.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/11_encoder_judge_prebuilts.ipynb) |
+| Performance metrics on a real, large-scale evaluation | [`12_performance_metrics_demo.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/12_performance_metrics_demo.ipynb) |
+| SGLang: environment checks, server mode via `api:`, tool calls and parallel tool calls | [`13_sglang_compatibility.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/13_sglang_compatibility.ipynb) |
+| Agent and RAG evals: tool calls, parallel tool calls, deployed agent endpoints, AgentTune data, retrieval and judged RAG metrics | [`14_agent_and_rag_evals.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/14_agent_and_rag_evals.ipynb) |
+| End-to-end agent evaluation, fully offline (`.py` script): recorded episodes, a verified state outcome, and a fabricated answer that does not pass a state case | [`agent_eval_offline.py`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/agent_eval_offline.py) |
 
 ## Applications
 
 Real-world scenarios answered end to end, not feature tours — see
-[`examples/applications/README.md`](examples/applications/README.md). Only `01` uses the
+[`examples/applications/README.md`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/applications/README.md). Only `01` uses the
 `lm-evaluation-harness` integration (`ak.run_lmeval()`, needs
 `auditkit[lmeval]`), as an authoritative cross-check alongside the native
 evaluation; none of the others do.
 
 | Application | File |
 |-------------|------|
-| How much does pruning severity (20%/40%/60%) degrade a model? Real BoolQ, real annotator, ship/no-ship verdicts, cross-checked via `ak.run_lmeval()` | [`01_application_pruned_llama_boolq.ipynb`](examples/applications/01_application_pruned_llama_boolq.ipynb) |
-| Healthcare: clinical QA correctness vs. grounding (PubMedQA) | [`02_application_healthcare_pubmedqa.ipynb`](examples/applications/02_application_healthcare_pubmedqa.ipynb) |
-| Finance: QA grounded in real SEC 10-K filings | [`03_application_finance_10k_qa.ipynb`](examples/applications/03_application_finance_10k_qa.ipynb) |
-| E-commerce: review-sentiment triage at scale | [`04_application_ecommerce_review_triage.ipynb`](examples/applications/04_application_ecommerce_review_triage.ipynb) |
-| Education: auto-graded tutoring, correctness vs. explanation | [`05_application_education_arc_tutor.ipynb`](examples/applications/05_application_education_arc_tutor.ipynb) |
-| Enterprise search: internal knowledge assistant (real retrieval + RAG grounding) | [`06_application_enterprise_search_rag.ipynb`](examples/applications/06_application_enterprise_search_rag.ipynb) |
-| LLM-as-judge via a real BERT NLI classifier, not a generative model | [`07_application_bert_nli_judge.ipynb`](examples/applications/07_application_bert_nli_judge.ipynb) |
+| How much does pruning severity (20%/40%/60%) degrade a model? Real BoolQ, real annotator, ship/no-ship verdicts, cross-checked via `ak.run_lmeval()` | [`01_application_pruned_llama_boolq.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/applications/01_application_pruned_llama_boolq.ipynb) |
+| Healthcare: clinical QA correctness vs. grounding (PubMedQA) | [`02_application_healthcare_pubmedqa.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/applications/02_application_healthcare_pubmedqa.ipynb) |
+| Finance: QA grounded in real SEC 10-K filings | [`03_application_finance_10k_qa.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/applications/03_application_finance_10k_qa.ipynb) |
+| E-commerce: review-sentiment triage at scale | [`04_application_ecommerce_review_triage.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/applications/04_application_ecommerce_review_triage.ipynb) |
+| Education: auto-graded tutoring, correctness vs. explanation | [`05_application_education_arc_tutor.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/applications/05_application_education_arc_tutor.ipynb) |
+| Enterprise search: internal knowledge assistant (real retrieval + RAG grounding) | [`06_application_enterprise_search_rag.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/applications/06_application_enterprise_search_rag.ipynb) |
+| LLM-as-judge via a real BERT NLI classifier, not a generative model | [`07_application_bert_nli_judge.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/applications/07_application_bert_nli_judge.ipynb) |
 
 ## Repository map
 
 | Directory | Description | README |
 |-----------|-------------|--------|
-| [`src/auditkit/`](src/auditkit/) | Core library — spine, runner, metrics, models, CLI | [README](src/auditkit/README.md) |
-| [`src/auditkit/metrics/`](src/auditkit/metrics/) | Metric families | [README](src/auditkit/metrics/README.md) |
-| [`src/auditkit/model/`](src/auditkit/model/) | Model backends (echo, openai, hf, vllm, etc.) | [README](src/auditkit/model/README.md) |
-| [`src/auditkit/redteam/`](src/auditkit/redteam/) | Red team probes and detectors | [README](src/auditkit/redteam/README.md) |
-| [`src/auditkit/scenarios/`](src/auditkit/scenarios/) | Built-in benchmark datasets | [README](src/auditkit/scenarios/README.md) |
-| [`tests/`](tests/) | Test suite (~3,000 tests) | [README](tests/README.md) |
-| [`examples/`](examples/) | Colab-ready example notebooks | [README](examples/README.md) |
-| [`examples/applications/`](examples/applications/) | Real-world application notebooks | [README](examples/applications/README.md) |
-| [`docs/`](docs/) | MkDocs source for [auditkit.lexsi.ai](https://auditkit.lexsi.ai/) | [index](docs/index.md) |
+| [`src/auditkit/`](https://github.com/Lexsi-Labs/AuditKIT/tree/main/src/auditkit/) | Core library — spine, runner, metrics, models, CLI | [README](https://github.com/Lexsi-Labs/AuditKIT/blob/main/src/auditkit/README.md) |
+| [`src/auditkit/metrics/`](https://github.com/Lexsi-Labs/AuditKIT/tree/main/src/auditkit/metrics/) | Metric families | [README](https://github.com/Lexsi-Labs/AuditKIT/blob/main/src/auditkit/metrics/README.md) |
+| [`src/auditkit/model/`](https://github.com/Lexsi-Labs/AuditKIT/tree/main/src/auditkit/model/) | Model backends (echo, openai, hf, vllm, etc.) | [README](https://github.com/Lexsi-Labs/AuditKIT/blob/main/src/auditkit/model/README.md) |
+| [`src/auditkit/redteam/`](https://github.com/Lexsi-Labs/AuditKIT/tree/main/src/auditkit/redteam/) | Red team probes and detectors | [README](https://github.com/Lexsi-Labs/AuditKIT/blob/main/src/auditkit/redteam/README.md) |
+| [`src/auditkit/scenarios/`](https://github.com/Lexsi-Labs/AuditKIT/tree/main/src/auditkit/scenarios/) | Built-in benchmark datasets | [README](https://github.com/Lexsi-Labs/AuditKIT/blob/main/src/auditkit/scenarios/README.md) |
+| [`tests/`](https://github.com/Lexsi-Labs/AuditKIT/tree/main/tests/) | Test suite (~3,000 tests) | [README](https://github.com/Lexsi-Labs/AuditKIT/blob/main/tests/README.md) |
+| [`examples/`](https://github.com/Lexsi-Labs/AuditKIT/tree/main/examples/) | Colab-ready example notebooks | [README](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/README.md) |
+| [`examples/applications/`](https://github.com/Lexsi-Labs/AuditKIT/tree/main/examples/applications/) | Real-world application notebooks | [README](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/applications/README.md) |
+| [`docs/`](https://github.com/Lexsi-Labs/AuditKIT/tree/main/docs/) | MkDocs source for [auditkit.lexsi.ai](https://auditkit.lexsi.ai/) | [index](https://github.com/Lexsi-Labs/AuditKIT/blob/main/docs/index.md) |
 
 ---
 
@@ -467,7 +469,7 @@ If you use AuditKIT in research, please cite it:
             Kadiyala, Ram Mohan Rao and Gosalia, Hem and Sankarapu, Vinay Kumar and
             Seth, Pratinav},
   year   = {2026},
-  version = {1.2.0},
+  version = {1.1.2},
   url    = {https://github.com/Lexsi-Labs/AuditKIT}
 }
 ```
@@ -478,11 +480,11 @@ If you use AuditKIT in research, please cite it:
 
 ## License
 
-This project is released under the [Lexsi Labs Source Available License (LSAL) v1.2](LICENSE.md) — free for academic research and teaching on MIT-like terms; use by any organization requires written acknowledgement or permission (Section 1A); a separate commercial license is required to sell it or embed it in a paid product.
+This project is released under the [Lexsi Labs Source Available License (LSAL) v1.2](https://github.com/Lexsi-Labs/AuditKIT/blob/main/LICENSE.md) — free for academic research and teaching on MIT-like terms; use by any organization requires written acknowledgement or permission (Section 1A); a separate commercial license is required to sell it or embed it in a paid product.
 
 ---
 
 ## Join Community / Contribute
 
 - Issues and discussions are welcomed on the [GitHub issue tracker](https://github.com/Lexsi-Labs/AuditKIT/issues).
-- See [Contributing](docs/community/contributing.md) for the development setup, code standards and the pull request process.
+- See [Contributing](https://github.com/Lexsi-Labs/AuditKIT/blob/main/docs/community/contributing.md) for the development setup, code standards and the pull request process.

{auditkit-1.1.0 → auditkit-1.1.2}/README.md

@@ -1,8 +1,8 @@
 <p align="center">
   <a href="https://github.com/Lexsi-Labs/AuditKIT">
     <picture>
-      <source media="(prefers-color-scheme: dark)" srcset="docs/assets/auditkit-logo-white.png">
-      <img src="docs/assets/auditkit-logo-black.png" alt="AuditKit" width="480">
+      <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/Lexsi-Labs/AuditKIT/main/docs/assets/auditkit-logo-white.png">
+      <img src="https://raw.githubusercontent.com/Lexsi-Labs/AuditKIT/main/docs/assets/auditkit-logo-black.png" alt="AuditKit" width="480">
     </picture>
   </a>
 </p>
@@ -14,7 +14,7 @@
 </p>
 
 <p align="center">
-  <a href="https://pypi.org/project/auditkit/"><img src="https://img.shields.io/badge/pypi-v1.2.0-0a8868" alt="PyPI v1.2.0"></a>
+  <a href="https://pypi.org/project/auditkit/"><img src="https://img.shields.io/pypi/v/auditkit?color=0a8868" alt="PyPI version"></a>
   <a href="https://www.python.org/"><img src="https://img.shields.io/badge/python-3.10%2B-blue" alt="Python 3.10+"></a>
   <a href="https://github.com/Lexsi-Labs/AuditKIT/blob/main/LICENSE.md"><img src="https://img.shields.io/badge/license-LSAL--1.2-lightgrey" alt="License: LSAL-1.2 (source-available, noncommercial)"></a>
   <a href="https://auditkit.lexsi.ai/"><img src="https://img.shields.io/badge/docs-auditkit.lexsi.ai-4c6ef5" alt="Documentation"></a>
@@ -40,7 +40,7 @@
 | **Models** | 10 backends: OpenAI, Anthropic, HuggingFace, Lexsi, vLLM, LiteLLM, API, Groq, OpenRouter, Agent (HTTP endpoint) — all resolved via `AutoModel.resolve()`. Any `list[str] → list[str]` callable also works. |
 | **Zero deps** | Core runs on stdlib. Backends and heavy metrics are optional extras (`pip install auditkit[openai]`). |
 | **Fingerprints** | Every run gets a stable sha256 — results are cacheable, comparable, and reproducible by construction. |
-| **Status** | ~3,000 tests, v1.2.0, LSAL-1.2 license (source-available, noncommercial). 11 metric families, 6 CLI subcommands, YAML config, MKDocs site. |
+| **Status** | ~3,000 tests, v1.1.2, LSAL-1.2 license (source-available, noncommercial). 11 metric families, 6 CLI subcommands, YAML config, MKDocs site. |
 
 ---
 
@@ -180,7 +180,7 @@ dropped it after 0.24.0, SGLang never had it), so both servers run it through th
 Transformers backends: automatically on vLLM, with `compat.py --patch-sglang` on SGLang. North Micro Vision is native on vLLM; on SGLang
 it needs transformers >= 5.15 in the SGLang env plus `--patch-sglang`. Tiny Aya and Aya Expanse
 are native on both. The full
-model-by-runtime table is in [docs/model_backends.md](docs/model_backends.md).
+model-by-runtime table is in [docs/model_backends.md](https://github.com/Lexsi-Labs/AuditKIT/blob/main/docs/model_backends.md).
 
 Tiny Aya (`CohereLabs/tiny-aya-global` and siblings `fire`, `water`, `earth`),
 Aya Expanse and Aya Vision are gated
@@ -213,57 +213,57 @@ Every `evaluate()` returns a `RunResult`:
 ## Examples
 
 Colab-ready notebooks with real models and real datasets — see
-[`examples/README.md`](examples/README.md) for the full, current list.
+[`examples/README.md`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/README.md) for the full, current list.
 
 | Example | File |
 |---------|------|
-| Full pipeline: adapter, 5 metrics, LLM judge, LLM annotator | [`01_full_evaluation_pipeline.ipynb`](examples/01_full_evaluation_pipeline.ipynb) |
-| Generation across two real HF model families, compared | [`02_generation_across_hf_families.ipynb`](examples/02_generation_across_hf_families.ipynb) |
-| Custom annotators (regex, LLM-backed, fully custom) | [`03_custom_annotators.ipynb`](examples/03_custom_annotators.ipynb) |
-| Metrics deep dive: built-in, custom, LLM-as-judge, RAG | [`04_metrics_deep_dive.ipynb`](examples/04_metrics_deep_dive.ipynb) |
-| Every data type/task kind: generative, MCQ, RAG, precomputed, chat | [`05_data_types.ipynb`](examples/05_data_types.ipynb) |
-| Model comparison deep dive: `compare_models()` + `RunComparison` | [`06_model_comparison.ipynb`](examples/06_model_comparison.ipynb) |
-| Annotators across 4 real model families, then compared together | [`07_annotators_across_models.ipynb`](examples/07_annotators_across_models.ipynb) |
-| `CompareResult` deep dive: what `compare_models()`'s native per-model support still can't express (different scorers per model), full method surface | [`08_compare_result_deep_dive.ipynb`](examples/08_compare_result_deep_dive.ipynb) |
-| `GuardJudge` implementation check: all 5 profiles, real and offline | [`09_guard_judge_implementation_check.ipynb`](examples/09_guard_judge_implementation_check.ipynb) |
-| `GuardJudge` with real, flagship guard models | [`10_guard_judge_legit_models.ipynb`](examples/10_guard_judge_legit_models.ipynb) |
-| `EncoderJudge`: the base mechanism plus its two prebuilt subclasses | [`11_encoder_judge_prebuilts.ipynb`](examples/11_encoder_judge_prebuilts.ipynb) |
-| Performance metrics on a real, large-scale evaluation | [`12_performance_metrics_demo.ipynb`](examples/12_performance_metrics_demo.ipynb) |
-| SGLang: environment checks, server mode via `api:`, tool calls and parallel tool calls | [`13_sglang_compatibility.ipynb`](examples/13_sglang_compatibility.ipynb) |
-| Agent and RAG evals: tool calls, parallel tool calls, deployed agent endpoints, AgentTune data, retrieval and judged RAG metrics | [`14_agent_and_rag_evals.ipynb`](examples/14_agent_and_rag_evals.ipynb) |
-| End-to-end agent evaluation, fully offline (`.py` script): recorded episodes, a verified state outcome, and a fabricated answer that does not pass a state case | [`agent_eval_offline.py`](examples/agent_eval_offline.py) |
+| Full pipeline: adapter, 5 metrics, LLM judge, LLM annotator | [`01_full_evaluation_pipeline.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/01_full_evaluation_pipeline.ipynb) |
+| Generation across two real HF model families, compared | [`02_generation_across_hf_families.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/02_generation_across_hf_families.ipynb) |
+| Custom annotators (regex, LLM-backed, fully custom) | [`03_custom_annotators.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/03_custom_annotators.ipynb) |
+| Metrics deep dive: built-in, custom, LLM-as-judge, RAG | [`04_metrics_deep_dive.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/04_metrics_deep_dive.ipynb) |
+| Every data type/task kind: generative, MCQ, RAG, precomputed, chat | [`05_data_types.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/05_data_types.ipynb) |
+| Model comparison deep dive: `compare_models()` + `RunComparison` | [`06_model_comparison.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/06_model_comparison.ipynb) |
+| Annotators across 4 real model families, then compared together | [`07_annotators_across_models.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/07_annotators_across_models.ipynb) |
+| `CompareResult` deep dive: what `compare_models()`'s native per-model support still can't express (different scorers per model), full method surface | [`08_compare_result_deep_dive.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/08_compare_result_deep_dive.ipynb) |
+| `GuardJudge` implementation check: all 5 profiles, real and offline | [`09_guard_judge_implementation_check.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/09_guard_judge_implementation_check.ipynb) |
+| `GuardJudge` with real, flagship guard models | [`10_guard_judge_legit_models.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/10_guard_judge_legit_models.ipynb) |
+| `EncoderJudge`: the base mechanism plus its two prebuilt subclasses | [`11_encoder_judge_prebuilts.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/11_encoder_judge_prebuilts.ipynb) |
+| Performance metrics on a real, large-scale evaluation | [`12_performance_metrics_demo.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/12_performance_metrics_demo.ipynb) |
+| SGLang: environment checks, server mode via `api:`, tool calls and parallel tool calls | [`13_sglang_compatibility.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/13_sglang_compatibility.ipynb) |
+| Agent and RAG evals: tool calls, parallel tool calls, deployed agent endpoints, AgentTune data, retrieval and judged RAG metrics | [`14_agent_and_rag_evals.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/14_agent_and_rag_evals.ipynb) |
+| End-to-end agent evaluation, fully offline (`.py` script): recorded episodes, a verified state outcome, and a fabricated answer that does not pass a state case | [`agent_eval_offline.py`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/agent_eval_offline.py) |
 
 ## Applications
 
 Real-world scenarios answered end to end, not feature tours — see
-[`examples/applications/README.md`](examples/applications/README.md). Only `01` uses the
+[`examples/applications/README.md`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/applications/README.md). Only `01` uses the
 `lm-evaluation-harness` integration (`ak.run_lmeval()`, needs
 `auditkit[lmeval]`), as an authoritative cross-check alongside the native
 evaluation; none of the others do.
 
 | Application | File |
 |-------------|------|
-| How much does pruning severity (20%/40%/60%) degrade a model? Real BoolQ, real annotator, ship/no-ship verdicts, cross-checked via `ak.run_lmeval()` | [`01_application_pruned_llama_boolq.ipynb`](examples/applications/01_application_pruned_llama_boolq.ipynb) |
-| Healthcare: clinical QA correctness vs. grounding (PubMedQA) | [`02_application_healthcare_pubmedqa.ipynb`](examples/applications/02_application_healthcare_pubmedqa.ipynb) |
-| Finance: QA grounded in real SEC 10-K filings | [`03_application_finance_10k_qa.ipynb`](examples/applications/03_application_finance_10k_qa.ipynb) |
-| E-commerce: review-sentiment triage at scale | [`04_application_ecommerce_review_triage.ipynb`](examples/applications/04_application_ecommerce_review_triage.ipynb) |
-| Education: auto-graded tutoring, correctness vs. explanation | [`05_application_education_arc_tutor.ipynb`](examples/applications/05_application_education_arc_tutor.ipynb) |
-| Enterprise search: internal knowledge assistant (real retrieval + RAG grounding) | [`06_application_enterprise_search_rag.ipynb`](examples/applications/06_application_enterprise_search_rag.ipynb) |
-| LLM-as-judge via a real BERT NLI classifier, not a generative model | [`07_application_bert_nli_judge.ipynb`](examples/applications/07_application_bert_nli_judge.ipynb) |
+| How much does pruning severity (20%/40%/60%) degrade a model? Real BoolQ, real annotator, ship/no-ship verdicts, cross-checked via `ak.run_lmeval()` | [`01_application_pruned_llama_boolq.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/applications/01_application_pruned_llama_boolq.ipynb) |
+| Healthcare: clinical QA correctness vs. grounding (PubMedQA) | [`02_application_healthcare_pubmedqa.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/applications/02_application_healthcare_pubmedqa.ipynb) |
+| Finance: QA grounded in real SEC 10-K filings | [`03_application_finance_10k_qa.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/applications/03_application_finance_10k_qa.ipynb) |
+| E-commerce: review-sentiment triage at scale | [`04_application_ecommerce_review_triage.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/applications/04_application_ecommerce_review_triage.ipynb) |
+| Education: auto-graded tutoring, correctness vs. explanation | [`05_application_education_arc_tutor.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/applications/05_application_education_arc_tutor.ipynb) |
+| Enterprise search: internal knowledge assistant (real retrieval + RAG grounding) | [`06_application_enterprise_search_rag.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/applications/06_application_enterprise_search_rag.ipynb) |
+| LLM-as-judge via a real BERT NLI classifier, not a generative model | [`07_application_bert_nli_judge.ipynb`](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/applications/07_application_bert_nli_judge.ipynb) |
 
 ## Repository map
 
 | Directory | Description | README |
 |-----------|-------------|--------|
-| [`src/auditkit/`](src/auditkit/) | Core library — spine, runner, metrics, models, CLI | [README](src/auditkit/README.md) |
-| [`src/auditkit/metrics/`](src/auditkit/metrics/) | Metric families | [README](src/auditkit/metrics/README.md) |
-| [`src/auditkit/model/`](src/auditkit/model/) | Model backends (echo, openai, hf, vllm, etc.) | [README](src/auditkit/model/README.md) |
-| [`src/auditkit/redteam/`](src/auditkit/redteam/) | Red team probes and detectors | [README](src/auditkit/redteam/README.md) |
-| [`src/auditkit/scenarios/`](src/auditkit/scenarios/) | Built-in benchmark datasets | [README](src/auditkit/scenarios/README.md) |
-| [`tests/`](tests/) | Test suite (~3,000 tests) | [README](tests/README.md) |
-| [`examples/`](examples/) | Colab-ready example notebooks | [README](examples/README.md) |
-| [`examples/applications/`](examples/applications/) | Real-world application notebooks | [README](examples/applications/README.md) |
-| [`docs/`](docs/) | MkDocs source for [auditkit.lexsi.ai](https://auditkit.lexsi.ai/) | [index](docs/index.md) |
+| [`src/auditkit/`](https://github.com/Lexsi-Labs/AuditKIT/tree/main/src/auditkit/) | Core library — spine, runner, metrics, models, CLI | [README](https://github.com/Lexsi-Labs/AuditKIT/blob/main/src/auditkit/README.md) |
+| [`src/auditkit/metrics/`](https://github.com/Lexsi-Labs/AuditKIT/tree/main/src/auditkit/metrics/) | Metric families | [README](https://github.com/Lexsi-Labs/AuditKIT/blob/main/src/auditkit/metrics/README.md) |
+| [`src/auditkit/model/`](https://github.com/Lexsi-Labs/AuditKIT/tree/main/src/auditkit/model/) | Model backends (echo, openai, hf, vllm, etc.) | [README](https://github.com/Lexsi-Labs/AuditKIT/blob/main/src/auditkit/model/README.md) |
+| [`src/auditkit/redteam/`](https://github.com/Lexsi-Labs/AuditKIT/tree/main/src/auditkit/redteam/) | Red team probes and detectors | [README](https://github.com/Lexsi-Labs/AuditKIT/blob/main/src/auditkit/redteam/README.md) |
+| [`src/auditkit/scenarios/`](https://github.com/Lexsi-Labs/AuditKIT/tree/main/src/auditkit/scenarios/) | Built-in benchmark datasets | [README](https://github.com/Lexsi-Labs/AuditKIT/blob/main/src/auditkit/scenarios/README.md) |
+| [`tests/`](https://github.com/Lexsi-Labs/AuditKIT/tree/main/tests/) | Test suite (~3,000 tests) | [README](https://github.com/Lexsi-Labs/AuditKIT/blob/main/tests/README.md) |
+| [`examples/`](https://github.com/Lexsi-Labs/AuditKIT/tree/main/examples/) | Colab-ready example notebooks | [README](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/README.md) |
+| [`examples/applications/`](https://github.com/Lexsi-Labs/AuditKIT/tree/main/examples/applications/) | Real-world application notebooks | [README](https://github.com/Lexsi-Labs/AuditKIT/blob/main/examples/applications/README.md) |
+| [`docs/`](https://github.com/Lexsi-Labs/AuditKIT/tree/main/docs/) | MkDocs source for [auditkit.lexsi.ai](https://auditkit.lexsi.ai/) | [index](https://github.com/Lexsi-Labs/AuditKIT/blob/main/docs/index.md) |
 
 ---
 
@@ -288,7 +288,7 @@ If you use AuditKIT in research, please cite it:
             Kadiyala, Ram Mohan Rao and Gosalia, Hem and Sankarapu, Vinay Kumar and
             Seth, Pratinav},
   year   = {2026},
-  version = {1.2.0},
+  version = {1.1.2},
   url    = {https://github.com/Lexsi-Labs/AuditKIT}
 }
 ```
@@ -299,11 +299,11 @@ If you use AuditKIT in research, please cite it:
 
 ## License
 
-This project is released under the [Lexsi Labs Source Available License (LSAL) v1.2](LICENSE.md) — free for academic research and teaching on MIT-like terms; use by any organization requires written acknowledgement or permission (Section 1A); a separate commercial license is required to sell it or embed it in a paid product.
+This project is released under the [Lexsi Labs Source Available License (LSAL) v1.2](https://github.com/Lexsi-Labs/AuditKIT/blob/main/LICENSE.md) — free for academic research and teaching on MIT-like terms; use by any organization requires written acknowledgement or permission (Section 1A); a separate commercial license is required to sell it or embed it in a paid product.
 
 ---
 
 ## Join Community / Contribute
 
 - Issues and discussions are welcomed on the [GitHub issue tracker](https://github.com/Lexsi-Labs/AuditKIT/issues).
-- See [Contributing](docs/community/contributing.md) for the development setup, code standards and the pull request process.
+- See [Contributing](https://github.com/Lexsi-Labs/AuditKIT/blob/main/docs/community/contributing.md) for the development setup, code standards and the pull request process.