devopsiq 0.1.0__tar.gz → 0.1.2__tar.gz

{devopsiq-0.1.0/devopsiq.egg-info → devopsiq-0.1.2}/PKG-INFO

@@ -1,12 +1,13 @@
 Metadata-Version: 2.4
 Name: devopsiq
-Version: 0.1.0
+Version: 0.1.2
 Summary: A from-scratch, read-only AI agent that investigates real DevOps problems — Kubernetes, Docker, Helm, Argo CD, Istio, Terraform, GitHub Actions, cloud, New Relic and more.
 Author: DevOpsAbhii
 License-Expression: MIT
 Project-URL: Homepage, https://github.com/DevOpsAbhii/devops-ai-agent
 Project-URL: Repository, https://github.com/DevOpsAbhii/devops-ai-agent
 Project-URL: Issues, https://github.com/DevOpsAbhii/devops-ai-agent/issues
+Project-URL: Changelog, https://github.com/DevOpsAbhii/devops-ai-agent/blob/main/CHANGELOG.md
 Keywords: devops,kubernetes,docker,helm,argocd,istio,terraform,github-actions,observability,newrelic,incident-response,root-cause-analysis,ai-agent,cli
 Classifier: Development Status :: 4 - Beta
 Classifier: Environment :: Console
@@ -29,6 +30,13 @@ Dynamic: license-file
 
 # DevOps AI Agent
 
+[![PyPI](https://img.shields.io/pypi/v/devopsiq)](https://pypi.org/project/devopsiq/)
+[![Python](https://img.shields.io/pypi/pyversions/devopsiq)](https://pypi.org/project/devopsiq/)
+[![Release](https://github.com/DevOpsAbhii/devops-ai-agent/actions/workflows/release.yml/badge.svg)](https://github.com/DevOpsAbhii/devops-ai-agent/actions/workflows/release.yml)
+[![Docker](https://img.shields.io/badge/ghcr-devops--ai--agent-2496ED?logo=docker&logoColor=white)](https://github.com/DevOpsAbhii/devops-ai-agent/pkgs/container/devops-ai-agent)
+[![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)
+[![Tests](https://img.shields.io/badge/tests-181%20offline-brightgreen)](https://github.com/DevOpsAbhii/devops-ai-agent/actions/workflows/release.yml)
+
 An AI agent that investigates real DevOps problems. The end goal: ask it
 something like *"Why is my Kubernetes pod in CrashLoopBackOff?"* and have it
 gather evidence, reason about the evidence, identify the likely root cause,
@@ -55,7 +63,14 @@ from environment config), and Ansible listing (inventory, playbook tasks).
 open alerts over NerdGraph — credentials from env), Trivy image
 vulnerability scanning, Helm releases (list/status/history), Argo CD
 (GitOps app sync/health), Istio mesh proxy status, and Docker Compose
-(project and service listing). The repository is git-tracked.
+(project and service listing). **Phase 10 ships it as a product:** the
+[`devopsiq` package on PyPI](https://pypi.org/project/devopsiq/) and a
+prebuilt multi-arch Docker image (`ghcr.io/devopsabhii/devops-ai-agent`) —
+both cut automatically by pushing a `v*` tag (test gate → PyPI via Trusted
+Publishing + GHCR in parallel). **No API key? The agent still runs:** it
+starts in model-less mode — record commands (`/report`, `/investigations`)
+and the whole 58-tool layer work without a key; only questions to the model
+need one. The repository is git-tracked.
 Every phase still built from scratch — no LangChain, LangGraph,
 AutoGen, CrewAI, or MCP.
 
@@ -73,7 +88,7 @@ model is called, how conversation history flows, how tool selection +
 execution + evidence feedback work — instead of depending on a framework
 for it.
 
-In Phase 5 the agent:
+Today the agent:
 - holds a conversation with **GLM 5.3** through **OpenRouter**;
 - has **58 real, read-only tools** across fifteen domains: host facts,
   Kubernetes (12 tools), Linux system (4), Docker + Compose (10),
@@ -133,8 +148,8 @@ Module map:
 
 | Path                        | Responsibility                                              |
 | --------------------------- | ------------------------------------------------------------ |
-| `main.py`                   | REPL loop, one-shot CLI, environment loading, `/investigate` commands, error messages |
-| `agent/agent.py`            | `DevOpsAgent` — client, history, `ask()`, `_complete()` loop |
+| `main.py`                   | REPL loop, one-shot CLI, environment loading, `/investigate` commands, error messages, model-less start |
+| `agent/agent.py`            | `DevOpsAgent` — client (`None` = model-less mode), history, `ask()`, `_complete()` loop |
 | `agent/prompts.py`          | The system prompt (versioned/tested separately)              |
 | `agent/investigation.py`    | First-class investigation record: hypotheses, verdicts, evidence, report renderer (pure data) |
 | `agent/store.py`            | `InvestigationStore` — one JSON file per record, atomic writes, resume/list (Phase 7) |
@@ -163,6 +178,11 @@ Module map:
 | `tests/test_phase7.py`      | Offline suite: round-trip serialization, store files, auto-save, resume, CLI flags (Phase 7) |
 | `tests/test_phase8.py`      | Offline suite: argv templates + validation for the 21 Phase 8 tools (fake CLIs, env-based monitoring) |
 | `tests/test_phase9.py`      | Offline suite: New Relic env-credential + payload tests, trivy/helm/argocd/istio/compose argv templates (Phase 9) |
+| `pyproject.toml`            | Package `devopsiq`: metadata, MIT, console script `devopsiq = main:main` (Phase 10) |
+| `Dockerfile`                | Prebuilt image: slim base + kubectl/helm/trivy/gh + app, non-root, `/data` record store (Phase 10) |
+| `.github/workflows/release.yml` | Tag-driven release: test gate → PyPI (Trusted Publishing) + GHCR multi-arch (Phase 10) |
+| `docs/generate_pdf.py`      | Builds the project documentation PDF from live source |
+| `docs/generate_release_guide.py` | Builds the release & update playbook PDF |
 
 ### The tool-use loop
 
@@ -209,6 +229,7 @@ the CLI exposes it directly:
 | `/investigations`        | List saved records on disk (newest first; `← active` marks the live one) |
 | `/report`                | Show the canonical report (once concluded)          |
 | `/endinvestigation`      | Clear the record (memory only — the saved copy stays as history) |
+| `/model [<name>]`        | Show the active model, or save a default (config.json) and switch to it |
 
 The report the agent ends with and `/report` render are kept consistent by
 construction: `tool` results confirm each record call and the tracker, and
@@ -396,9 +417,39 @@ official `openai` Python SDK works as our client with two config lines:
 OpenAI(api_key=..., base_url="https://openrouter.ai/api/v1")
 ```
 
-Swapping to another OpenRouter model later is a one-line change (an env var
-today); moving to any other OpenAI-compatible provider changes only
-`agent/agent.py` configuration.
+Swapping to another OpenRouter model later is a one-liner (env var, `/model`
+command, or `--model` flag); moving to any other OpenAI-compatible provider
+changes only `agent/agent.py` configuration.
+
+**Your model, your choice.** The default is baked in as a fallback, never a
+restriction. Four ways to set the model — the first one set wins:
+
+1. `--model NAME` flag (one run: `devopsiq --model openai/gpt-5.2 "why?"`)
+2. `/model <name>` in the REPL — saves it to `~/.devops-ai-agent/config.json`
+   so every future run uses it
+3. `OPENROUTER_MODEL` (shell or `.env`)
+4. built-in default (`z-ai/glm-5.3`)
+
+```bash
+export OPENROUTER_MODEL=anthropic/claude-sonnet-5   # or openai/gpt-5.2, google/gemini-2.5-pro, ...
+devopsiq /model openai/gpt-5.2                      # or save a default from the REPL
+```
+
+Two things to weigh when picking: the agent is a tool-use loop, so choose a
+model with solid function-calling (a chat-only model will answer from
+imagination instead of gathering evidence); and an investigation makes
+several model calls per run, so price-per-call multiplies — that is why the
+default is a cheap, reliable tool-caller rather than the biggest model.
+
+**No OpenRouter at all?** `OPENROUTER_BASE_URL` points the same client at
+any OpenAI-compatible endpoint — including a local one. Ollama, free and
+offline:
+
+```bash
+export OPENROUTER_BASE_URL=http://localhost:11434/v1
+export OPENROUTER_MODEL=llama3.2          # any Ollama model that does tools
+devopsiq "docker container checkout keeps exiting, investigate"
+```
 
 ## 5. Installation
 
@@ -463,6 +514,27 @@ Recommended CLIs, by domain:
 | New Relic    | `curl` + `NEW_RELIC_API_KEY` / `NEW_RELIC_ACCOUNT_ID` env vars |
 | Ansible      | `ansible-inventory`, `ansible-playbook`          |
 
+### No API key? You still have the tools
+
+The model is the only part that needs a key — the agent builds without one
+(model-less mode), and two things keep working:
+
+- **Record commands**: `devopsiq /report`, `devopsiq /investigations`, and
+  the same commands inside the REPL, all work with no key configured. Only
+  actual questions exit 1 with the setup message naming `OPENROUTER_API_KEY`.
+- **The whole 58-tool layer** via the Python library, with no model and no
+  key (see INTEGRATION.md):
+
+```python
+import agent.agent                            # registers every tool
+from tools.registry import execute_tool
+
+print(execute_tool("k8s_pods", '{"namespace": "prod"}'))
+```
+
+And the fully key-free *agent* path is a local model (Ollama example in
+section 4 above) — no cloud account needed at all.
+
 ## 6. Environment setup
 
 ```bash
@@ -490,7 +562,9 @@ NEW_RELIC_ACCOUNT_ID=1234567                        # (Phase 9) numeric account
 
 `.env` is gitignored; the API key is never hard-coded in Python, printed, or
 logged. If `OPENROUTER_API_KEY` is already set in your shell, the shell value
-wins and `.env` is not consulted. Kubernetes tools use kubectl's own config
+wins and `.env` is not consulted. Without any key the agent still starts in
+model-less mode (record commands + the tool layer — see "No API key?" in
+section 5). Kubernetes tools use kubectl's own config
 (`~/.kube/config` or `KUBECONFIG`); no agent-side config is needed.
 
 ## 7. How to run the agent
@@ -511,6 +585,7 @@ setup/API errors — so it drops straight into a pipeline:
 .venv/bin/python main.py --json "why is api-5d6f crash-looping?"   # structured report
 .venv/bin/python main.py --resume --json "any update?"             # continue a prior run
 .venv/bin/python main.py --store-dir /tmp/runs --out report.json "..."  # pipeline paths
+.venv/bin/python main.py --model openai/gpt-5.2 "why is it down?"   # one-run model override
 ```
 
 With `--json` the stdout is one JSON document (see `render_report_json` in
@@ -552,8 +627,8 @@ Tools:   ansible_inventory, ansible_playbook_tasks, aws_identity,
          k8s_top_nodes, k8s_top_pods, loki_query, prom_query,
          sys_open_ports, sys_service_logs, sys_service_status,
          sys_top_processes, system_info, tf_plan, tf_show, tf_state_list
-Commands: /investigate <problem>, /investigation, /investigations, /report, /endinvestigation
-One-shot: python main.py [--json] [--resume] [--out report.json] [--store-dir DIR] "<problem>"
+Commands: /investigate <problem>, /investigation, /investigations, /report, /endinvestigation, /model [<name>]
+One-shot: python main.py [--json] [--resume] [--out report.json] [--store-dir DIR] [--model NAME] "<problem>"
 Type 'exit' to quit.
 
 You: The checkout service container keeps exiting in Docker. Investigate.
@@ -570,7 +645,7 @@ You: exit
 Type `exit` / `quit`, or press Ctrl-D / Ctrl-C to leave. Slash commands are
 handled locally and never reach the model.
 
-## 8. Current limitations (Phase 9)
+## 8. Current limitations
 
 - **Each domain needs its CLI installed and reachable.** Missing CLIs,
   unauthenticated `gh`, a dead docker daemon, an uninitialized terraform
@@ -655,8 +730,22 @@ handled locally and never reach the model.
   `helm_history` — reads only), Argo CD (`argocd_apps`,
   `argocd_app_status`), Istio (`istioctl_proxy_status`), Docker Compose
   (`docker_compose_ls`, `docker_compose_ps` — live-verified on this host).
+- **Phase 10 — distribution. ✅ Done.** The `devopsiq` package on PyPI
+  (console command `devopsiq`, MIT) and the prebuilt multi-arch GHCR image,
+  cut by a tag-driven release workflow: an offline test gate, then PyPI
+  (Trusted Publishing — no token in the repo) and GHCR (amd64 + arm64) in
+  parallel. **`v0.1.1` added model-less mode:** with no API key the agent
+  still constructs — record commands (`/report`, `/investigations`) and the
+  58-tool layer work; only model questions exit 1 with the setup message.
+  Verified end-to-end against the published package.
+- **Model choice + resilient calls (v0.1.2). ✅ Done.** The model is
+  set by a 4-rung ladder — `--model` flag > `/model`-saved config file
+  (`~/.devops-ai-agent/config.json`) > `OPENROUTER_MODEL` > baked-in
+  default — and transient model-call failures (timeouts, connection
+  errors, 429s, 5xx) retry with exponential backoff (2s→4s→8s + jitter,
+  honoring a 429's Retry-After); a rejected key fails immediately.
 - **Later — region-scoped cloud resources** (ec2 describe-*, compute
   instances list, ...) behind the same template pattern; more observability
   depth (New Relic dashboards/entities, Prometheus range queries); streaming;
-  conversation-history persistence; and a human-approval gate before any
-  mutating action is ever allowed.
+  conversation-history persistence; and a human-approval gate
+  before any mutating action is ever allowed.

{devopsiq-0.1.0 → devopsiq-0.1.2}/README.md

@@ -1,5 +1,12 @@
 # DevOps AI Agent
 
+[![PyPI](https://img.shields.io/pypi/v/devopsiq)](https://pypi.org/project/devopsiq/)
+[![Python](https://img.shields.io/pypi/pyversions/devopsiq)](https://pypi.org/project/devopsiq/)
+[![Release](https://github.com/DevOpsAbhii/devops-ai-agent/actions/workflows/release.yml/badge.svg)](https://github.com/DevOpsAbhii/devops-ai-agent/actions/workflows/release.yml)
+[![Docker](https://img.shields.io/badge/ghcr-devops--ai--agent-2496ED?logo=docker&logoColor=white)](https://github.com/DevOpsAbhii/devops-ai-agent/pkgs/container/devops-ai-agent)
+[![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)
+[![Tests](https://img.shields.io/badge/tests-181%20offline-brightgreen)](https://github.com/DevOpsAbhii/devops-ai-agent/actions/workflows/release.yml)
+
 An AI agent that investigates real DevOps problems. The end goal: ask it
 something like *"Why is my Kubernetes pod in CrashLoopBackOff?"* and have it
 gather evidence, reason about the evidence, identify the likely root cause,
@@ -26,7 +33,14 @@ from environment config), and Ansible listing (inventory, playbook tasks).
 open alerts over NerdGraph — credentials from env), Trivy image
 vulnerability scanning, Helm releases (list/status/history), Argo CD
 (GitOps app sync/health), Istio mesh proxy status, and Docker Compose
-(project and service listing). The repository is git-tracked.
+(project and service listing). **Phase 10 ships it as a product:** the
+[`devopsiq` package on PyPI](https://pypi.org/project/devopsiq/) and a
+prebuilt multi-arch Docker image (`ghcr.io/devopsabhii/devops-ai-agent`) —
+both cut automatically by pushing a `v*` tag (test gate → PyPI via Trusted
+Publishing + GHCR in parallel). **No API key? The agent still runs:** it
+starts in model-less mode — record commands (`/report`, `/investigations`)
+and the whole 58-tool layer work without a key; only questions to the model
+need one. The repository is git-tracked.
 Every phase still built from scratch — no LangChain, LangGraph,
 AutoGen, CrewAI, or MCP.
 
@@ -44,7 +58,7 @@ model is called, how conversation history flows, how tool selection +
 execution + evidence feedback work — instead of depending on a framework
 for it.
 
-In Phase 5 the agent:
+Today the agent:
 - holds a conversation with **GLM 5.3** through **OpenRouter**;
 - has **58 real, read-only tools** across fifteen domains: host facts,
   Kubernetes (12 tools), Linux system (4), Docker + Compose (10),
@@ -104,8 +118,8 @@ Module map:
 
 | Path                        | Responsibility                                              |
 | --------------------------- | ------------------------------------------------------------ |
-| `main.py`                   | REPL loop, one-shot CLI, environment loading, `/investigate` commands, error messages |
-| `agent/agent.py`            | `DevOpsAgent` — client, history, `ask()`, `_complete()` loop |
+| `main.py`                   | REPL loop, one-shot CLI, environment loading, `/investigate` commands, error messages, model-less start |
+| `agent/agent.py`            | `DevOpsAgent` — client (`None` = model-less mode), history, `ask()`, `_complete()` loop |
 | `agent/prompts.py`          | The system prompt (versioned/tested separately)              |
 | `agent/investigation.py`    | First-class investigation record: hypotheses, verdicts, evidence, report renderer (pure data) |
 | `agent/store.py`            | `InvestigationStore` — one JSON file per record, atomic writes, resume/list (Phase 7) |
@@ -134,6 +148,11 @@ Module map:
 | `tests/test_phase7.py`      | Offline suite: round-trip serialization, store files, auto-save, resume, CLI flags (Phase 7) |
 | `tests/test_phase8.py`      | Offline suite: argv templates + validation for the 21 Phase 8 tools (fake CLIs, env-based monitoring) |
 | `tests/test_phase9.py`      | Offline suite: New Relic env-credential + payload tests, trivy/helm/argocd/istio/compose argv templates (Phase 9) |
+| `pyproject.toml`            | Package `devopsiq`: metadata, MIT, console script `devopsiq = main:main` (Phase 10) |
+| `Dockerfile`                | Prebuilt image: slim base + kubectl/helm/trivy/gh + app, non-root, `/data` record store (Phase 10) |
+| `.github/workflows/release.yml` | Tag-driven release: test gate → PyPI (Trusted Publishing) + GHCR multi-arch (Phase 10) |
+| `docs/generate_pdf.py`      | Builds the project documentation PDF from live source |
+| `docs/generate_release_guide.py` | Builds the release & update playbook PDF |
 
 ### The tool-use loop
 
@@ -180,6 +199,7 @@ the CLI exposes it directly:
 | `/investigations`        | List saved records on disk (newest first; `← active` marks the live one) |
 | `/report`                | Show the canonical report (once concluded)          |
 | `/endinvestigation`      | Clear the record (memory only — the saved copy stays as history) |
+| `/model [<name>]`        | Show the active model, or save a default (config.json) and switch to it |
 
 The report the agent ends with and `/report` render are kept consistent by
 construction: `tool` results confirm each record call and the tracker, and
@@ -367,9 +387,39 @@ official `openai` Python SDK works as our client with two config lines:
 OpenAI(api_key=..., base_url="https://openrouter.ai/api/v1")
 ```
 
-Swapping to another OpenRouter model later is a one-line change (an env var
-today); moving to any other OpenAI-compatible provider changes only
-`agent/agent.py` configuration.
+Swapping to another OpenRouter model later is a one-liner (env var, `/model`
+command, or `--model` flag); moving to any other OpenAI-compatible provider
+changes only `agent/agent.py` configuration.
+
+**Your model, your choice.** The default is baked in as a fallback, never a
+restriction. Four ways to set the model — the first one set wins:
+
+1. `--model NAME` flag (one run: `devopsiq --model openai/gpt-5.2 "why?"`)
+2. `/model <name>` in the REPL — saves it to `~/.devops-ai-agent/config.json`
+   so every future run uses it
+3. `OPENROUTER_MODEL` (shell or `.env`)
+4. built-in default (`z-ai/glm-5.3`)
+
+```bash
+export OPENROUTER_MODEL=anthropic/claude-sonnet-5   # or openai/gpt-5.2, google/gemini-2.5-pro, ...
+devopsiq /model openai/gpt-5.2                      # or save a default from the REPL
+```
+
+Two things to weigh when picking: the agent is a tool-use loop, so choose a
+model with solid function-calling (a chat-only model will answer from
+imagination instead of gathering evidence); and an investigation makes
+several model calls per run, so price-per-call multiplies — that is why the
+default is a cheap, reliable tool-caller rather than the biggest model.
+
+**No OpenRouter at all?** `OPENROUTER_BASE_URL` points the same client at
+any OpenAI-compatible endpoint — including a local one. Ollama, free and
+offline:
+
+```bash
+export OPENROUTER_BASE_URL=http://localhost:11434/v1
+export OPENROUTER_MODEL=llama3.2          # any Ollama model that does tools
+devopsiq "docker container checkout keeps exiting, investigate"
+```
 
 ## 5. Installation
 
@@ -434,6 +484,27 @@ Recommended CLIs, by domain:
 | New Relic    | `curl` + `NEW_RELIC_API_KEY` / `NEW_RELIC_ACCOUNT_ID` env vars |
 | Ansible      | `ansible-inventory`, `ansible-playbook`          |
 
+### No API key? You still have the tools
+
+The model is the only part that needs a key — the agent builds without one
+(model-less mode), and two things keep working:
+
+- **Record commands**: `devopsiq /report`, `devopsiq /investigations`, and
+  the same commands inside the REPL, all work with no key configured. Only
+  actual questions exit 1 with the setup message naming `OPENROUTER_API_KEY`.
+- **The whole 58-tool layer** via the Python library, with no model and no
+  key (see INTEGRATION.md):
+
+```python
+import agent.agent                            # registers every tool
+from tools.registry import execute_tool
+
+print(execute_tool("k8s_pods", '{"namespace": "prod"}'))
+```
+
+And the fully key-free *agent* path is a local model (Ollama example in
+section 4 above) — no cloud account needed at all.
+
 ## 6. Environment setup
 
 ```bash
@@ -461,7 +532,9 @@ NEW_RELIC_ACCOUNT_ID=1234567                        # (Phase 9) numeric account
 
 `.env` is gitignored; the API key is never hard-coded in Python, printed, or
 logged. If `OPENROUTER_API_KEY` is already set in your shell, the shell value
-wins and `.env` is not consulted. Kubernetes tools use kubectl's own config
+wins and `.env` is not consulted. Without any key the agent still starts in
+model-less mode (record commands + the tool layer — see "No API key?" in
+section 5). Kubernetes tools use kubectl's own config
 (`~/.kube/config` or `KUBECONFIG`); no agent-side config is needed.
 
 ## 7. How to run the agent
@@ -482,6 +555,7 @@ setup/API errors — so it drops straight into a pipeline:
 .venv/bin/python main.py --json "why is api-5d6f crash-looping?"   # structured report
 .venv/bin/python main.py --resume --json "any update?"             # continue a prior run
 .venv/bin/python main.py --store-dir /tmp/runs --out report.json "..."  # pipeline paths
+.venv/bin/python main.py --model openai/gpt-5.2 "why is it down?"   # one-run model override
 ```
 
 With `--json` the stdout is one JSON document (see `render_report_json` in
@@ -523,8 +597,8 @@ Tools:   ansible_inventory, ansible_playbook_tasks, aws_identity,
          k8s_top_nodes, k8s_top_pods, loki_query, prom_query,
          sys_open_ports, sys_service_logs, sys_service_status,
          sys_top_processes, system_info, tf_plan, tf_show, tf_state_list
-Commands: /investigate <problem>, /investigation, /investigations, /report, /endinvestigation
-One-shot: python main.py [--json] [--resume] [--out report.json] [--store-dir DIR] "<problem>"
+Commands: /investigate <problem>, /investigation, /investigations, /report, /endinvestigation, /model [<name>]
+One-shot: python main.py [--json] [--resume] [--out report.json] [--store-dir DIR] [--model NAME] "<problem>"
 Type 'exit' to quit.
 
 You: The checkout service container keeps exiting in Docker. Investigate.
@@ -541,7 +615,7 @@ You: exit
 Type `exit` / `quit`, or press Ctrl-D / Ctrl-C to leave. Slash commands are
 handled locally and never reach the model.
 
-## 8. Current limitations (Phase 9)
+## 8. Current limitations
 
 - **Each domain needs its CLI installed and reachable.** Missing CLIs,
   unauthenticated `gh`, a dead docker daemon, an uninitialized terraform
@@ -626,8 +700,22 @@ handled locally and never reach the model.
   `helm_history` — reads only), Argo CD (`argocd_apps`,
   `argocd_app_status`), Istio (`istioctl_proxy_status`), Docker Compose
   (`docker_compose_ls`, `docker_compose_ps` — live-verified on this host).
+- **Phase 10 — distribution. ✅ Done.** The `devopsiq` package on PyPI
+  (console command `devopsiq`, MIT) and the prebuilt multi-arch GHCR image,
+  cut by a tag-driven release workflow: an offline test gate, then PyPI
+  (Trusted Publishing — no token in the repo) and GHCR (amd64 + arm64) in
+  parallel. **`v0.1.1` added model-less mode:** with no API key the agent
+  still constructs — record commands (`/report`, `/investigations`) and the
+  58-tool layer work; only model questions exit 1 with the setup message.
+  Verified end-to-end against the published package.
+- **Model choice + resilient calls (v0.1.2). ✅ Done.** The model is
+  set by a 4-rung ladder — `--model` flag > `/model`-saved config file
+  (`~/.devops-ai-agent/config.json`) > `OPENROUTER_MODEL` > baked-in
+  default — and transient model-call failures (timeouts, connection
+  errors, 429s, 5xx) retry with exponential backoff (2s→4s→8s + jitter,
+  honoring a 429's Retry-After); a rejected key fails immediately.
 - **Later — region-scoped cloud resources** (ec2 describe-*, compute
   instances list, ...) behind the same template pattern; more observability
   depth (New Relic dashboards/entities, Prometheus range queries); streaming;
-  conversation-history persistence; and a human-approval gate before any
-  mutating action is ever allowed.
+  conversation-history persistence; and a human-approval gate
+  before any mutating action is ever allowed.

{devopsiq-0.1.0 → devopsiq-0.1.2}/agent/agent.py

@@ -26,9 +26,18 @@ exits; the delegates below expose resume/list/store-location to the CLI.
 """
 
 import os
+import random
+import time
+
+from openai import (
+    APIConnectionError,
+    APIStatusError,
+    APITimeoutError,
+    OpenAI,
+    RateLimitError,
+)
 
-from openai import OpenAI
-
+from agent.config import load_user_config
 from agent.prompts import SYSTEM_PROMPT
 from agent.store import InvestigationStore
 from tools import (  # noqa: F401 — side effect: each module registers its tools
@@ -68,6 +77,50 @@ PLACEHOLDER_KEY = "your_key_here"
 # Safety valve: the model gets at most this many tool-use turns before we stop.
 MAX_TOOL_ITERATIONS = 10
 
+# Transient-failure retries for the model call itself (network blips, 429,
+# 5xx). MAX_MODEL_RETRIES attempts beyond the first, exponential backoff
+# starting at RETRY_BASE_DELAY (2s -> 4s -> 8s) plus jitter, capped at
+# RETRY_MAX_DELAY. A 429's Retry-After header wins when present.
+MAX_MODEL_RETRIES = 3
+RETRY_BASE_DELAY = 2.0
+RETRY_MAX_DELAY = 60.0
+
+_sleep = time.sleep  # test seam: offline tests patch agent.agent._sleep
+
+
+def _is_transient(exc: Exception) -> bool:
+    """True when retrying `exc` can plausibly succeed.
+
+    Timeouts, connection failures, rate limits, and server-side 5xx are
+    transient. A rejected key (401) or any other client 4xx is not — the
+    same request would fail identically forever, so it fails immediately.
+    """
+    if isinstance(exc, (APITimeoutError, APIConnectionError, RateLimitError)):
+        return True
+    if isinstance(exc, APIStatusError):
+        return exc.status_code >= 500
+    return False
+
+
+def _retry_delay(exc: Exception, attempt: int) -> float:
+    """Seconds to wait before retry number `attempt` (0-based).
+
+    Exponential backoff with jitter, except on a rate limit carrying a
+    Retry-After header — the server knows its own budget, so it wins
+    (clamped to [1, RETRY_MAX_DELAY] so a bad header cannot hurt us).
+    """
+    response = getattr(exc, "response", None)
+    retry_after = getattr(response, "headers", {}).get("retry-after")
+    if retry_after:
+        try:
+            return min(max(float(retry_after), 1.0), RETRY_MAX_DELAY)
+        except (TypeError, ValueError):
+            pass  # non-numeric header: fall through to backoff
+    return min(
+        RETRY_BASE_DELAY * (2 ** attempt) + random.uniform(0, 1),
+        RETRY_MAX_DELAY,
+    )
+
 
 class DevOpsAgent:
     """Minimal DevOps investigation assistant (read-only by design)."""
@@ -80,23 +133,35 @@ class DevOpsAgent:
     ) -> None:
         # Configuration resolution order: explicit argument > environment > default.
         self.api_key = api_key or os.getenv("OPENROUTER_API_KEY")
+
+        self.model = (
+            model                                   # 1. explicit --model flag
+            or load_user_config().get("model")      # 2. ~/.devops-ai-agent/config.json
+            or os.getenv("OPENROUTER_MODEL")        # 3. environment / .env
+            or DEFAULT_MODEL                        # 4. built-in default
+        )
+        self.base_url = base_url or os.getenv("OPENROUTER_BASE_URL", DEFAULT_BASE_URL)
+
+        # Model-less mode: a missing (or placeholder) key no longer blocks
+        # construction. Record operations (the slash commands) and the tool
+        # layer never call the model, so the agent builds with client=None
+        # and only the chat path (_complete) demands a working key.
+        self.client = None
+        self.model_error: str | None = None
         if not self.api_key:
-            raise ValueError(
+            self.model_error = (
                 "OPENROUTER_API_KEY is not set. Copy .env.example to .env and "
                 "fill in your key, or export OPENROUTER_API_KEY in your shell."
             )
-        if self.api_key.strip().lower() == PLACEHOLDER_KEY:
-            raise ValueError(
+        elif self.api_key.strip().lower() == PLACEHOLDER_KEY:
+            self.model_error = (
                 "OPENROUTER_API_KEY still has the placeholder value. Edit .env "
                 "and replace `your_key_here` with your real key."
             )
-
-        self.model = model or os.getenv("OPENROUTER_MODEL", DEFAULT_MODEL)
-        self.base_url = base_url or os.getenv("OPENROUTER_BASE_URL", DEFAULT_BASE_URL)
-
-        # OpenRouter exposes an OpenAI-compatible API, so the official OpenAI
-        # SDK is a drop-in client — just pointed at OpenRouter's base URL.
-        self.client = OpenAI(api_key=self.api_key, base_url=self.base_url)
+        else:
+            # OpenRouter exposes an OpenAI-compatible API, so the official
+            # OpenAI SDK is a drop-in client — just pointed at the base URL.
+            self.client = OpenAI(api_key=self.api_key, base_url=self.base_url)
 
         # Short-term conversation history. Starts with the system prompt;
         # grows as user, model, and tool results exchange turns.
@@ -110,7 +175,8 @@ class DevOpsAgent:
 
         The message, any tool exchanges, and the final assistant reply are all
         kept in this session's history so the model retains context across
-        turns. Raises ValueError on empty input.
+        turns. Raises ValueError on empty input, or when no API key is
+        configured (model-less mode).
         """
         message = user_message.strip()
         if not message:
@@ -180,12 +246,16 @@ class DevOpsAgent:
         model again. Stops when the model answers in plain text, or when the
         iteration cap is reached.
         """
+        if self.client is None:
+            # Model-less construction: the first chat attempt is where the
+            # missing key finally surfaces, as a normal caught ValueError.
+            raise ValueError(self.model_error)
         for _ in range(MAX_TOOL_ITERATIONS):
             request: dict = {"model": self.model, "messages": self.messages}
             if self.tools:
                 request["tools"] = [tool.schema() for tool in self.tools]
 
-            response = self.client.chat.completions.create(**request)
+            response = self._call_model(request)
             message = response.choices[0].message
 
             if not message.tool_calls:
@@ -207,6 +277,23 @@ class DevOpsAgent:
             f"The model did not finish after {MAX_TOOL_ITERATIONS} tool-use turns."
         )
 
+    def _call_model(self, request: dict):
+        """One model call with bounded retries on transient failures.
+
+        Timeouts, connection errors, rate limits (429) and server-side 5xx
+        are retried up to MAX_MODEL_RETRIES times with exponential backoff
+        (2s, 4s, 8s + jitter); a 429's Retry-After header wins when present.
+        Client errors — a rejected key (401) most notably — fail immediately,
+        because retrying the identical request cannot fix them.
+        """
+        for attempt in range(MAX_MODEL_RETRIES + 1):
+            try:
+                return self.client.chat.completions.create(**request)
+            except Exception as exc:  # noqa: BLE001 — classified right below
+                if attempt >= MAX_MODEL_RETRIES or not _is_transient(exc):
+                    raise
+                _sleep(_retry_delay(exc, attempt))
+
 
 def _echo_tool_request(message) -> dict:
     """Rebuild the assistant turn that requested tools, verbatim.

devopsiq-0.1.2/agent/config.py

@@ -0,0 +1,45 @@
+"""Per-user preferences: ~/.devops-ai-agent/config.json.
+
+The `devopsiq` equivalent of "my settings" — currently the preferred model,
+written by the /model command and read by the agent at startup. Precedence
+for the model, highest first:
+
+    --model flag  >  config.json  >  OPENROUTER_MODEL env  >  built-in default
+
+The file is tiny JSON, one flat object. Every read is tolerant: a missing
+file, an unreadable home directory, or corrupt content yields {} — a broken
+config never blocks the agent (the same degrade-don't-crash rule the
+investigation store follows). Tests point AGENT_CONFIG_FILE at a tmp path.
+"""
+
+import json
+import os
+from pathlib import Path
+
+DEFAULT_CONFIG_PATH = Path.home() / ".devops-ai-agent" / "config.json"
+
+
+def config_path() -> Path:
+    """The active config file (AGENT_CONFIG_FILE overrides; tests use tmp)."""
+    return Path(os.getenv("AGENT_CONFIG_FILE") or DEFAULT_CONFIG_PATH)
+
+
+def load_user_config() -> dict:
+    """Read the config file; {} when missing or unreadable — never raise."""
+    try:
+        raw = config_path().read_text(encoding="utf-8")
+        data = json.loads(raw)
+        return data if isinstance(data, dict) else {}
+    except (OSError, ValueError):
+        return {}
+
+
+def save_user_config(update: dict) -> Path:
+    """Merge `update` into the config file (create the directory if needed)."""
+    path = config_path()
+    data = load_user_config()
+    data.update(update)
+    path.parent.mkdir(parents=True, exist_ok=True)
+    path.write_text(json.dumps(data, indent=2, sort_keys=True) + "\n",
+                    encoding="utf-8")
+    return path