docs/content/usage.md (1)

30-32: Align unordered list style to asterisk to satisfy markdownlint (MD004).

Earlier lists in this doc use asterisks; these three use dashes, which triggers lint. Consistent style also improves readability.

- - **Model Experimentation**: Easy switching between Ollama, LM Studio and OpenAI backends  
- - **Resource Management**: Automatic failover when local resources are busy
- - **Cost Optimisation**: Priority routing (local first, cloud fallback via [LiteLLM](compare/litellm.md))
+ * **Model Experimentation**: Easy switching between Ollama, LM Studio and OpenAI backends  
+ * **Resource Management**: Automatic failover when local resources are busy
+ * **Cost Optimisation**: Priority routing (local first, cloud fallback via [LiteLLM](compare/litellm.md))

docs/content/compare/localai.md (6)

33-33: Add a language identifier to the fenced code block (MD040).

Specifying a language silences lint and improves rendering. For ASCII diagrams, use “text”.

-```
+```text

---

118-118: Add a language identifier to the fenced code block (MD040).

Same rationale as above; use “text” for ASCII diagrams.

-```
+```text

---

67-67: Remove trailing punctuation in headings (MD026).

Colons at the end of headings trigger lint and don’t add value.

-### Use Olla When:
+### Use Olla When

---

75-75: Remove trailing punctuation in headings (MD026).

-### Use LocalAI When:
+### Use LocalAI When

---

107-107: Remove trailing punctuation in headings (MD026).

-### Benefits:
+### Benefits

---

196-199: Qualify latency claims to avoid overprecision and environment-dependence.

Hard numbers can be misleading across environments. Suggest softening to “typical/approximate” to reflect variability.

-- **Direct to LocalAI**: Baseline
-- **Through Olla**: +2ms routing overhead
-- **Benefit**: Faster failover than timeout/retry
+- **Direct to LocalAI**: Baseline
+- **Through Olla**: ~2 ms routing overhead (typical; varies by environment)
+- **Benefit**: Faster failover than client-side timeout/retry in most deployments

If you have recent benchmarks, consider linking them here for transparency.

docs/content/compare/overview.md (3)

57-57: Add language identifiers to fenced code blocks (MD040).

Use “text” for ASCII diagrams to satisfy lint and improve rendering.

-```
+```text

---

68-68: Add language identifiers to fenced code blocks (MD040).

-```
+```text

---

79-79: Add language identifiers to fenced code blocks (MD040).

-```
+```text

docs/content/compare/litellm.md (6)

33-33: Add language identifier to fenced code block (MD040).

ASCII architecture diagrams should use “text” for lint compliance.

-```
+```text

---

41-41: Add language identifier to fenced code block (MD040).

-```
+```text

---

116-116: Add language identifier to fenced code block (MD040).

-```
+```text

---

74-74: Remove trailing punctuation in headings (MD026).

-### Use Olla When:
+### Use Olla When

---

82-82: Remove trailing punctuation in headings (MD026).

-### Use LiteLLM When:
+### Use LiteLLM When

---

156-161: Qualify latency figures to reflect environmental variance.

Numbers like “<2ms” and “10–50ms” vary by hardware/network. Softening the language avoids misleading readers.

-- **Olla alone**: <2ms overhead
-- **LiteLLM alone**: 10-50ms overhead
-- **Olla + LiteLLM**: ~12-52ms total overhead
+- **Olla alone**: typically <2 ms overhead
+- **LiteLLM alone**: typically 10–50 ms overhead
+- **Olla + LiteLLM**: ~12–52 ms total overhead (typical)

If you have internal measurements, link or footnote them for credibility.

docs/content/compare/gpustack.md (7)

33-33: Add language identifier to fenced code block (MD040).

Use “text” for the stack-position ASCII diagram.

-```
+```text

---

74-74: Remove trailing punctuation in headings (MD026).

-### Use Olla When:
+### Use Olla When

---

82-82: Remove trailing punctuation in headings (MD026).

-### Use GPUStack When:
+### Use GPUStack When

---

87-87: Fix minor verb agreement.

Reads more naturally as singular in this bullet list.

-- Require automatic model distribution
+- Requires automatic model distribution

---

136-136: Add language identifier to fenced code block (MD040).

Use “text” for the ASCII diagram.

-```
+```text

---

153-153: Add language identifier to fenced code block (MD040).

Use “text” for the ASCII diagram.

-```
+```text

---

163-163: Add language identifier to fenced code block (MD040).

Use “text” for the ASCII diagram.

-```
+```text

docs/content/compare/integration-patterns.md (7)

9-9: Tighten phrasing (“missing article”)

Minor copy edit for readability in en‑AU: add an article before “robust LLM infrastructure”.

-This guide shows how to combine Olla with other tools to build robust LLM infrastructure for different use cases.
+This guide shows how to combine Olla with other tools to build a robust LLM infrastructure for different use cases.

---

17-24: Add a language to fenced block to satisfy markdownlint (MD040)

These ASCII diagrams are fenced without a language. Mark as text to fix linting and improve rendering.

-```
+```text
 Before:  Apps → LLM Endpoints (single point of failure)

 After:   Apps → Olla → LLM Endpoints (automatic failover)
                    ├── Primary endpoint
                    ├── Secondary endpoint
                    └── Tertiary endpoint

---

`50-55`: **Add a language to fenced block to satisfy markdownlint (MD040)**

Same issue here; use “text”.


```diff
-```
+```text
          Olla
            ├── Local GPU (priority 1)
            ├── [LiteLLM](./litellm.md) → Cloud APIs (priority 10)
            └── [LocalAI](./localai.md)/[Ollama](https://github.com/ollama/ollama) (priority 2)

---

`85-89`: **Add a language to fenced block to satisfy markdownlint (MD040)**

Marking this as “text” keeps the monospace diagram formatting while passing linting.


```diff
-```
+```text
 Team A Apps → Olla Config A → Team A Resources
 Team B Apps → Olla Config B → Shared Resources + Team B
 Production  → Olla Config C → Production Pool

---

`103-108`: **Add a language to fenced block to satisfy markdownlint (MD040)**

Mark these region diagrams as “text”.


```diff
-```
+```text
      Global Olla
          ├── Sydney [GPUStack](./gpustack.md) (for ANZ users)
          ├── Singapore [LocalAI](./localai.md) (for APAC users)
          └── US [vLLM](https://github.com/vllm-project/vllm) (for Americas users)

---

`328-345`: **Use the GHCR image and pin to a version for reproducibility**

The Quick Start elsewhere uses GHCR. Align the compose example and pin to a version tag to avoid “latest” drift.


```diff
   olla:
-    image: thushan/olla:latest
+    image: ghcr.io/thushan/olla:vX.Y.Z
     ports:
       - "8080:8080"
     volumes:
       - ./config.yaml:/config.yaml
@@
   litellm:
-    image: ghcr.io/berriai/litellm:latest
+    image: ghcr.io/berriai/litellm:vA.B.C

If you’d prefer to keep “latest”, consider at least documenting the tested versions above the block.

---

352-362: Healthcheck command may require curl inside the container

Many minimal images don’t include curl. Consider a CMD-SHELL form with wget or switching to a small helper sidecar.

   healthcheck:
-      test: ["CMD", "curl", "-f", "http://localhost:8080/internal/health"]
+      test: ["CMD-SHELL", "wget -qO- http://localhost:8080/internal/health >/dev/null 2>&1 || exit 1"]
       interval: 10s

docs/mkdocs.yml (1)

119-123: Fix indentation for nested nav items and standardise casing

YAML lint flagged indent; children under “Compare” should be further indented. Also use “GPUStack” casing consistently with the rest of the docs.

-      - Compare:
-        - Overview: compare/overview.md
-        - Patterns: compare/integration-patterns.md
-        - vs GpuStack: compare/gpustack.md
-        - vs LiteLLM: compare/litellm.md
-        - vs LocalAI: compare/localai.md
+      - Compare:
+          - Overview: compare/overview.md
+          - Patterns: compare/integration-patterns.md
+          - vs GPUStack: compare/gpustack.md
+          - vs LiteLLM: compare/litellm.md
+          - vs LocalAI: compare/localai.md

docs/content/index.md (1)

27-28: Neutral, clearer phrasing for positioning (en‑AU)

Minor copy tweak to avoid “Unlike … like” phrasing and keep the emphasis on reliability.

-Unlike API gateways like [LiteLLM](compare/litellm.md) or orchestration platforms like [GPUStack](compare/gpustack.md), Olla focuses on making your existing LLM infrastructure reliable through intelligent routing and failover.
+Compared to API gateways such as [LiteLLM](compare/litellm.md) or orchestration platforms such as [GPUStack](compare/gpustack.md), Olla focuses on improving the reliability of your existing LLM infrastructure through intelligent routing and failover.

docs/content/faq.md (2)

194-209: Clarify API key expectations when proxying “OpenAI-compatible”

Great example. Consider a brief note that if an upstream (e.g. LiteLLM or LocalAI) enforces auth, an API key may still be required even when talking to Olla.

 Yes, Olla provides OpenAI-compatible endpoints (similar to [LocalAI](compare/localai.md)):
@@
 response = client.chat.completions.create(
     model="llama3.2",
     messages=[{"role": "user", "content": "Hello"}]
 )
+
+Note: When Olla proxies to a backend that enforces authentication (e.g. a secured LiteLLM or LocalAI instance), you’ll still need to provide a valid API key expected by that upstream.

---

210-213: Reinforce “works together” with a direct link to patterns

Nice positioning. Add a pointer to concrete integration patterns.

-[LiteLLM](compare/litellm.md) is an API translation layer for cloud providers, while Olla is an infrastructure proxy for self-hosted endpoints. They solve different problems and work well together - LiteLLM for cloud APIs, Olla for local infrastructure reliability.
+[LiteLLM](compare/litellm.md) is an API translation layer for cloud providers, while Olla is an infrastructure proxy for self-hosted endpoints. They solve different problems and work well together — LiteLLM for cloud APIs, Olla for local infrastructure reliability. See [integration patterns](compare/integration-patterns.md#tool-specific-integrations) for concrete setups.

Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro

💡 Knowledge Base configuration:

• MCP integration is disabled by default for public repositories
• Jira integration is disabled by default for public repositories
• Linear integration is disabled by default for public repositories

You can enable these sources in your CodeRabbit configuration.