lio-lang 0.1.0__tar.gz

lio_lang-0.1.0/LICENSE

@@ -0,0 +1,200 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for describing the origin of the Work and
+      reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may accept support,
+      warranty, indemnity, or other liability obligations and/or rights
+      consistent with this License. However, in accepting such obligations,
+      You may act only on behalf of and on Your own responsibility, not on
+      behalf of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright 2026 Estevao Fonseca
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

lio_lang-0.1.0/NOTICE

@@ -0,0 +1,5 @@
+lio-lang
+Copyright 2026 Estevao Fonseca
+
+This product includes software developed by Estevao Fonseca
+(https://github.com/estevaofon/lio-lang).

lio_lang-0.1.0/PKG-INFO

@@ -0,0 +1,492 @@
+Metadata-Version: 2.4
+Name: lio-lang
+Version: 0.1.0
+Summary: Structured key:value protocol for machine-to-machine LLM communication
+Author: Estevao
+License-Expression: Apache-2.0
+Project-URL: Homepage, https://github.com/estevaofon/lio-lang
+Project-URL: Repository, https://github.com/estevaofon/lio-lang
+Project-URL: Issues, https://github.com/estevaofon/lio-lang/issues
+Project-URL: Documentation, https://github.com/estevaofon/lio-lang#readme
+Keywords: llm,prompt-engineering,agents,anthropic,openai,claude
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Developers
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3 :: Only
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Software Development :: Libraries :: Python Modules
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Typing :: Typed
+Requires-Python: >=3.11
+Description-Content-Type: text/markdown
+License-File: LICENSE
+License-File: NOTICE
+Provides-Extra: anthropic
+Requires-Dist: anthropic>=0.40.0; extra == "anthropic"
+Provides-Extra: benchmarks
+Requires-Dist: anthropic>=0.40.0; extra == "benchmarks"
+Requires-Dist: python-dotenv>=1.0.0; extra == "benchmarks"
+Requires-Dist: pydantic>=2.0; extra == "benchmarks"
+Provides-Extra: dev
+Requires-Dist: pytest>=8.0; extra == "dev"
+Requires-Dist: anthropic>=0.40.0; extra == "dev"
+Requires-Dist: python-dotenv>=1.0.0; extra == "dev"
+Requires-Dist: pydantic>=2.0; extra == "dev"
+Dynamic: license-file
+
+# lio-lang
+
+Protocolo estruturado key:value para comunicacao maquina-a-maquina com LLMs.
+Substitui a prosa que o modelo produz entre tool calls e em respostas
+finais por um formato compacto parecido com JSON, voltado para outputs
+que serao consumidos por codigo, agentes ou automacao em vez de lidos
+por humanos.
+
+```python
+import lio
+
+prompt = lio.system_prompt(mode="agentic", language="en")
+report = lio.validate(model_text)           # compliance check
+natural = lio.render(lio_text)               # LIO -> natural for display
+metrics = lio.measure(                       # cost/savings math
+    natural_input_tokens=582, natural_output_tokens=25378,
+    lio_input_tokens=4311, lio_output_tokens=5488,
+    model="claude-opus-4-6",
+)
+```
+
+Sem dependencias obrigatorias. Licenca Apache 2.0. Python 3.11+.
+
+---
+
+## Instalacao
+
+```bash
+pip install lio-lang                    # core, zero dependencias
+pip install lio-lang[anthropic]         # com wrapper da SDK Anthropic
+pip install lio-lang[benchmarks]        # com deps para rodar os benchmarks
+
+# Ou com uv
+uv sync
+uv sync --extra anthropic
+uv sync --extra benchmarks
+```
+
+---
+
+## Quickstart
+
+```python
+import anthropic
+import lio
+
+client = anthropic.Anthropic()
+
+response = client.messages.create(
+    model="claude-opus-4-6",
+    max_tokens=16000,
+    system=lio.system_prompt(mode="single_shot", language="en"),
+    messages=[{"role": "user", "content": "Write a Python prime filter function"}],
+)
+
+raw_text = "".join(b.text for b in response.content if b.type == "text")
+
+# Check the response is compliant LIO
+report = lio.validate(raw_text)
+print(f"compliant: {report.compliant} (score {report.score})")
+
+# Convert LIO to natural language for human display (no extra API call)
+print(lio.render(raw_text, language="en"))
+```
+
+Exemplos rodaveis em `examples/`:
+
+```bash
+# Exercita toda a API publica, funciona offline e online
+uv run examples/quickstart.py                       # sem API key: so offline
+uv run --extra anthropic examples/quickstart.py     # com API key: online tambem
+
+# Compara baseline natural vs LIO roundtrip lado a lado
+uv run --extra anthropic examples/full_flow.py --prompt "Faca uma funcao de ordenacao"
+uv run --extra anthropic examples/full_flow.py --prompt "Write a debounce function" --language en
+```
+
+---
+
+## API publica
+
+| Funcao | Proposito |
+|---|---|
+| `lio.system_prompt(mode, language)` | Retorna a string do system prompt LIO. `mode="agentic"` (multi-turno com tools) ou `"single_shot"`. `language="en"` ou `"pt"`. |
+| `lio.validate(text)` | Checa se um text block e LIO valido. Retorna `ComplianceReport(compliant, score, reasons)`. |
+| `lio.render(lio_text, language)` | Converte LIO de volta para linguagem natural para display. Rule-based, zero chamadas de API. Fallback seguro retorna o input inalterado. |
+| `lio.measure(...)` | Calcula economia de custo a partir de contagens pareadas de tokens. Retorna `EconomyMetrics` com custo, economia percentual, breakdown de input/output. |
+| `lio.middleware.wrap_call` | Decorator que injeta o system prompt LIO em qualquer funcao de chamada LLM e opcionalmente roda um callback de telemetria. |
+
+### Exemplo de middleware
+
+```python
+import lio
+from lio.middleware import wrap_call
+
+def telemetry(response):
+    text = "".join(b.text for b in response.content if b.type == "text")
+    r = lio.validate(text)
+    print(f"[lio] compliance={r.score:.2f}")
+
+@wrap_call(mode="agentic", language="en", on_response=telemetry)
+def call_claude(messages, **kwargs):
+    return client.messages.create(
+        model="claude-opus-4-6",
+        max_tokens=16000,
+        messages=messages,
+        **kwargs,
+    )
+
+# Cada chamada vai automaticamente:
+#  1. receber o system prompt LIO injetado
+#  2. ser validada via callback
+response = call_claude(messages=[{"role": "user", "content": "..."}])
+```
+
+---
+
+## Filosofia
+
+**LIO estrutura informacao, nao comprime.**
+
+O objetivo nao e "economizar tokens a qualquer custo". E transferir a
+mesma informacao que linguagem natural transferiria, eliminando o
+overhead de apresentacao: prosa, markdown, pedagogia, cortesia,
+decoracao visual.
+
+O que cai fora sao as sobras de linguagem natural, nao os dados.
+Detalhes tecnicos que perderiam funcionalidade se abreviados
+(paths absolutos, identificadores exatos, valores de config, schemas)
+sao preservados. Quando a resposta precisa de valores que o consumidor
+vai substituir, usa-se placeholders explicitos `<REPLACE:nome>` em vez
+de tentar comprimir com truques que colidem com sintaxe da linguagem
+alvo.
+
+A analogia correta e com JSON, nao com compressao lossy:
+
+> LIO esta para prosa assim como JSON esta para XML-verboso: mesma
+> informacao, sem a gordura de apresentacao. JSON nao e "compressao
+> lossy de XML"; e uma estruturacao diferente da mesma informacao.
+> LIO tem a mesma relacao com a prosa de LLM.
+
+Consequencias praticas:
+
+- Uma resposta LIO pode ser longa se a informacao exige.
+- Use quantos campos forem necessarios para manter precisao tecnica.
+- Alternativas validas sao preservadas via campo `variant`.
+- O modelo mantem liberdade de planejamento, nao e constrangido a
+  "escolher uma resposta e descartar as outras".
+
+---
+
+## Sintaxe
+
+### Gramatica minima
+
+```
+field        = key ':' value
+value        = escalar | '[' lista ']' | '{' objeto '}' | '<REPLACE:' nome '>'
+text_block   = sequencia de fields, separados por espaco ou newline
+```
+
+Escalares sao literais. Listas usam virgula. Objetos usam virgula entre pares.
+
+### Campos reservados
+
+Uma letra para a estrutura base:
+
+| Campo | Uso |
+|---|---|
+| `T` | task type |
+| `O` | object / target |
+| `P` | params |
+| `C` | context |
+| `F` | expected format |
+| `X` | constraints |
+| `S` | status (`ok`, `err`, `partial`) |
+| `R` | result |
+| `D` | data |
+| `W` | warning |
+| `E` | error |
+| `Q` | query |
+| `N` | count |
+
+Palavras-chave para semantica adicional:
+
+| Campo | Uso |
+|---|---|
+| `step` | acao atual (loops agenticos) |
+| `next` | proxima acao |
+| `fix` | correcao proposta |
+| `issues` | lista de problemas encontrados |
+| `causes` | lista de causas possiveis |
+| `variant` | resposta alternativa valida |
+| `note` | observacao nao obvia |
+| `code` | bloco de codigo (passa intacto) |
+| `steps` | sequencia de passos |
+
+Voce pode introduzir campos novos quando fizer sentido.
+
+### Placeholders explicitos
+
+Quando a resposta precisa de valores que o consumidor deve substituir,
+use `<REPLACE:nome_descritivo>`. Por que nao outras opcoes:
+
+- `$variavel` ou `${variavel}` colidem com bash, nginx, make
+- `{{variavel}}` colide com Jinja, Mustache, Go templates
+- `__variavel__` fragil, nao se destaca
+- `<REPLACE:nome>` angle brackets sao slot universal, `REPLACE` e
+  explicito, nenhuma linguagem mainstream usa esse token literal
+
+Exemplo (config nginx com SSL):
+
+```
+S:ok R:nginx_config
+code:{
+server {
+    listen 443 ssl http2;
+    server_name <REPLACE:your_domain>;
+    ssl_certificate /etc/letsencrypt/live/<REPLACE:your_domain>/fullchain.pem;
+    ssl_certificate_key /etc/letsencrypt/live/<REPLACE:your_domain>/privkey.pem;
+    location / { proxy_pass http://localhost:3000; }
+}
+}
+W:run_certbot_first->certbot --nginx -d <REPLACE:your_domain>
+```
+
+### O que LIO nao tem
+
+O mesmo que JSON nao tem: markdown headers, tabelas, listas bullet,
+emoji decorativo, blocos de citacao, negrito/italico, preambulo
+("Aqui esta..."), cortesia ("Espero que ajude"), ou prosa explicativa
+redundante com o conteudo.
+
+Codigo dentro de `code:{}` passa intacto, pode conter qualquer sintaxe
+da linguagem alvo.
+
+---
+
+## Quando usar LIO
+
+Use onde o consumidor da resposta **nao e um humano**, ou onde a
+resposta sera processada por codigo antes de chegar em humano.
+
+**Casos ideais:**
+- Pipelines agentic (agente LLM agente) - nenhum humano le o intermediario
+- Workflows automatizados (n8n, Make, Zapier)
+- RAG chains - query, retrieve, LLM, resposta estruturada downstream
+- Multi-agent systems onde agentes conversam entre si
+- Batch processing - milhares de chamadas identicas em formato
+- Coding agent plugins - raciocinio intermediario entre tool calls
+
+**Nao use para:**
+- Chat direto com humano - o humano precisa ler a resposta
+- Tarefas criativas (escrita, brainstorm) - a verbosidade e o produto
+- Educacao e explicacoes - tokens de explicacao sao o produto
+- Respostas onde a prosa **e** o valor (emails, documentacao)
+
+---
+
+## Padroes de integracao
+
+### 1. Injecao direta do system prompt
+
+O mais simples. Funciona com qualquer SDK:
+
+```python
+import anthropic
+import lio
+
+client = anthropic.Anthropic()
+response = client.messages.create(
+    model="claude-opus-4-6",
+    max_tokens=16000,
+    system=lio.system_prompt(mode="agentic", language="en"),
+    messages=[{"role": "user", "content": "..."}],
+)
+```
+
+### 2. Middleware decorator
+
+Se voce ja tem uma funcao que chama a LLM, so decora:
+
+```python
+import lio
+from lio.middleware import wrap_call
+
+@wrap_call(mode="agentic", language="en", on_response=lambda r: print("[lio] ok"))
+def call_llm(messages, **kwargs):
+    return client.messages.create(
+        model="claude-opus-4-6",
+        max_tokens=16000,
+        messages=messages,
+        **kwargs,
+    )
+
+response = call_llm(messages=[{"role": "user", "content": "..."}])
+```
+
+### 3. Renderizacao para UI
+
+Se o consumidor final e humano, converta LIO de volta para natural
+antes de mostrar:
+
+```python
+text = "".join(b.text for b in response.content if b.type == "text")
+human_view = lio.render(text, language="en")
+print(human_view)
+```
+
+Transformacao rule-based local, sem chamada de API. Cobre os padroes
+comuns (`S:ok code:{...}`, `issues:[...] fix:[...]`, objetos aninhados,
+listas) e faz fallback seguro retornando o input inalterado quando nao
+reconhece a forma.
+
+---
+
+## Benchmarks
+
+Bateria de 15 tarefas single-shot contra `claude-opus-4-6`, pareando
+prompts em linguagem natural com prompts em LIO. Roda com:
+
+```bash
+python -m benchmarks.run                 # 15 tarefas, ~$1
+python -m benchmarks.multi_turn          # sessao agentic, ~$0.15
+python -m benchmarks.fidelity            # LLM-as-judge, ~$0.30
+```
+
+### Resultados representativos
+
+| Cenario | Compliance | Fidelidade | Economia custo | Economia output |
+|---|---|---|---|---|
+| Single-shot, 15 tarefas | 100% | 95% | 69-75% | 79-82% |
+| Multi-turn agentic, 3 turnos | 100% | - | 40-55% | 55-73% |
+| Roundtrip Fibonacci (example full_flow) | 100% | - | 70-75% | 83-87% |
+
+Compliance e medido pelo validator local (`lio.validate`). Fidelidade
+e medida com LLM-as-judge calibrado em `benchmarks/fidelity.py`.
+Economia de output varia por categoria:
+
+| Categoria | Economia de output |
+|---|---|
+| debug | ~70% |
+| cfg | ~85% |
+| refactor | ~85% |
+| code | ~80% |
+| analyze | ~65% |
+
+Economia efetiva varia com a proporcao de text vs tool_use no output
+total. Para coding agents com uso pesado de tools, a economia por
+sessao tipicamente fica entre 30-50%, pois LIO so afeta os text blocks
+do modelo - tool_use blocks JSON permanecem identicos.
+
+---
+
+## Limites conhecidos
+
+1. **Economia depende do shape da tarefa.** Tarefas prose-heavy
+   (explicacoes, analises, debug) capturam 60-75%. Tarefas
+   tool-heavy (coding agents com muito read/write/bash) capturam
+   30-50%. Tarefas onde o output e dominado por codigo cru dentro
+   de `code:{}` capturam 20-30% (o codigo passa intacto).
+
+2. **Input savings sao frequentemente negativos.** O system prompt
+   LIO adiciona ~500 tokens por chamada. Em prompts curtos isso
+   domina o input. A economia vem do output, que custa 5x mais
+   que input no Opus. Em sessoes longas com historico acumulado,
+   input savings vira positivo.
+
+3. **LIO requer modelos inteligentes.** Testado em Opus 4.6 e
+   Sonnet 4.6. Nao recomendado para Haiku - o overhead cognitivo
+   de seguir o protocolo nao compensa para modelos menores.
+   Haiku pode ser usado como decoder LIO->natural barato.
+
+4. **Drift de compliance em prompts-gatilho.** Palavras como
+   "relatorio"/"report" no prompt do usuario ativam priors fortes
+   para output em markdown. O validator detecta drift (inclui
+   detecao de markdown inclusive dentro de `code:{}` blocks), e a
+   filosofia "estrutura nao compressao" mitiga, mas vale atencao.
+
+5. **Fidelidade depende de juiz calibrado.** LLM-as-judge pode
+   penalizar LIO por "omissoes" que sao na verdade over-delivery
+   da resposta natural (info que o usuario nao pediu). O juiz em
+   `benchmarks/fidelity.py` e calibrado para ancorar o julgamento
+   na pergunta do usuario, nao na resposta natural de referencia.
+
+---
+
+## Estrutura do repositorio
+
+```
+lio-lang/
+├── lio/
+│   ├── __init__.py               # API publica
+│   ├── protocol.py               # factory de system prompts
+│   ├── validator.py              # compliance checker
+│   ├── render.py                 # LIO -> natural (rule-based)
+│   ├── metrics.py                # calculo de economia + tabela de pricing
+│   ├── client.py                 # wrapper opcional da Anthropic SDK
+│   ├── languages/
+│   │   ├── en.py                 # prompts em ingles
+│   │   └── pt.py                 # prompts em portugues
+│   └── middleware/
+│       └── generic.py            # wrap_call decorator
+├── tests/
+│   ├── test_validator.py
+│   ├── test_render.py
+│   └── test_metrics.py
+├── benchmarks/
+│   ├── tasks.jsonl               # 15 tarefas pareadas
+│   ├── run.py                    # benchmark single-shot
+│   ├── multi_turn.py             # benchmark agentic
+│   ├── fidelity.py               # LLM-as-judge
+│   └── compare.py                # relatorio comparativo
+├── examples/
+│   ├── quickstart.py             # exercita toda a API (offline + online)
+│   └── full_flow.py              # comparacao natural vs LIO roundtrip
+├── pyproject.toml
+├── LICENSE                       # MIT
+└── README.md                     # este arquivo
+```
+
+---
+
+## Desenvolvimento
+
+```bash
+git clone https://github.com/estevaofon/lio-lang
+cd lio-lang
+uv sync --extra dev              # ou: pip install -e .[dev]
+
+# Rodar os testes
+pytest tests/
+
+# Rodar os exemplos (offline, sem API key)
+uv run examples/quickstart.py --offline
+
+# Rodar os exemplos com API real (precisa ANTHROPIC_API_KEY em .env)
+uv run --extra anthropic examples/quickstart.py
+uv run --extra anthropic examples/full_flow.py --prompt "sua pergunta"
+
+# Rodar benchmark completo (custo ~$1.50 em API)
+uv run --extra benchmarks python -m benchmarks.run
+uv run --extra benchmarks python -m benchmarks.multi_turn
+uv run --extra benchmarks python -m benchmarks.fidelity
+```
+
+---
+
+## Licenca
+
+Apache License 2.0. Ver `LICENSE` e `NOTICE`.
+
+Copyright 2026 Estevao Fonseca.