dm: stop creating a duplicate DM room on every restart

GetDMRoom's lookup chain could never hit. The in-memory cache died with the
process, and the m.direct account data it fell back on was never written by
the bot -- room creation only ever populated the map. As an appservice user
there is no /sync to populate it either, so after each restart the first DM to
a user fell through both checks and created a fresh room.

Persist the mapping in a new dm_rooms table and read it before anything else.
Creation now also publishes m.direct, so clients label the room as a DM.

A stored room is checked for liveness before reuse, counting the user as
present while merely invited -- they often never accept, and treating that as
gone would recreate the room on every send, which is the original bug wearing
a different hat. The same check now guards the m.direct path, which previously
took the last entry on faith.

Two adoption paths keep existing rooms from being orphaned: user-initiated DM
invites are claimed on join, and a room the user is already talking in is
adopted on first message (negative-cached so group rooms cost one member
lookup, not one per message). For users predating the table, a one-shot sweep
of joined rooms runs only on the path that would otherwise create a duplicate.

The sweep cannot rank duplicates -- without /sync there are no timestamps, so
it takes the newest by room-list order. First message from the user corrects
any wrong pick.

internal/db/db.go

@@ -1255,6 +1255,16 @@ CREATE TABLE IF NOT EXISTS presence (
 	updated_at INTEGER DEFAULT (unixepoch())
 );
 
+-- DM rooms. The bot's m.direct account data is not a reliable store for an
+-- appservice user (no /sync, and nothing writes it back), so the mapping lives
+-- here. Without it every restart lost the cache and the bot created a fresh DM
+-- room per user.
+CREATE TABLE IF NOT EXISTS dm_rooms (
+	user_id TEXT PRIMARY KEY,
+	room_id TEXT NOT NULL,
+	updated_at INTEGER DEFAULT (unixepoch())
+);
+
 -- Markov
 CREATE TABLE IF NOT EXISTS markov_corpus (
 	id INTEGER PRIMARY KEY AUTOINCREMENT,

internal/llm/llm.go

@@ -0,0 +1,161 @@
+// Package llm wraps the local inference endpoint behind a backend-agnostic
+// interface. Two concrete backends — Ollama (native /api/generate) and vLLM
+// (OpenAI-compatible /v1/chat/completions) — implement Client; plugin code
+// calls the interface only and never knows which one is active.
+//
+// Deliberately not routed through internal/safehttp: that client blocks
+// RFC1918 and loopback destinations to defend against SSRF from feed-supplied
+// URLs, and the inference endpoint is precisely such a destination. The URL
+// here comes from our own config, never from user input.
+package llm
+
+import (
+	"context"
+	"os"
+	"strings"
+	"time"
+)
+
+// Request is the backend-neutral generation request. Zero-valued fields fall
+// back to backend defaults.
+type Request struct {
+	// Prompt is a single raw instruction. The vLLM backend wraps it as one
+	// user message so the model's chat template still applies; sending it to
+	// /v1/completions instead would bypass the template and degrade an
+	// instruction-tuned model badly.
+	Prompt string
+	// System is an optional system message. Empty means none, which keeps the
+	// single-message shape the majority of callers use.
+	System string
+	// NumCtx is the per-request context window. Ollama honours it directly;
+	// vLLM fixes the window server-side at launch (--max-model-len), so this
+	// is ignored there rather than silently misapplied.
+	NumCtx int
+	// MaxTokens caps the completion length. 0 means the backend default.
+	MaxTokens int
+	// Temperature is passed through when non-zero.
+	Temperature float64
+	// Timeout overrides the client's default per-request budget.
+	Timeout time.Duration
+}
+
+// Client is the single surface plugin code depends on.
+type Client interface {
+	// Generate returns the full completion in one shot. Reasoning blocks are
+	// stripped before returning (see StripThink) — every caller in this repo
+	// wants the visible answer, not the chain of thought.
+	Generate(ctx context.Context, req Request) (string, error)
+	// Model reports the configured model id, for logging and /botinfo.
+	Model() string
+	// Ping reports the model ids the backend is currently serving. Used by
+	// /botinfo for a liveness line; the two backends expose this on different
+	// paths (/api/tags vs /v1/models), which is exactly the sort of difference
+	// this interface exists to hide.
+	Ping(ctx context.Context) ([]string, error)
+	// Backend reports "ollama" or "vllm", for logging and /botinfo.
+	Backend() string
+}
+
+// Config selects and configures a backend.
+type Config struct {
+	Backend  string // "ollama" | "vllm"
+	Endpoint string
+	Model    string
+	Timeout  time.Duration
+}
+
+// DefaultTimeout matches the budget the pre-refactor callOllama used.
+const DefaultTimeout = 120 * time.Second
+
+// ConfigFromEnv reads backend settings, preferring the new LLM_* names and
+// falling back to the legacy OLLAMA_* pair so an existing deployment keeps
+// working untouched after this refactor.
+func ConfigFromEnv() Config {
+	backend := strings.ToLower(strings.TrimSpace(os.Getenv("LLM_BACKEND")))
+	if backend == "" {
+		backend = "ollama"
+	}
+
+	endpoint := firstNonEmpty(os.Getenv("LLM_ENDPOINT"), os.Getenv("OLLAMA_HOST"))
+	model := firstNonEmpty(os.Getenv("LLM_MODEL"), os.Getenv("OLLAMA_MODEL"))
+
+	timeout := DefaultTimeout
+	if d, err := time.ParseDuration(os.Getenv("LLM_TIMEOUT")); err == nil && d > 0 {
+		timeout = d
+	}
+
+	return Config{Backend: backend, Endpoint: endpoint, Model: model, Timeout: timeout}
+}
+
+// Configured reports whether enough config is present to talk to a backend.
+// Plugins check this to stay dormant rather than erroring on every invocation,
+// which is what the old `if ollamaHost == "" || ollamaModel == ""` guards did.
+func (c Config) Configured() bool {
+	return c.Endpoint != "" && c.Model != ""
+}
+
+// New builds the client for cfg.Backend. An unrecognised backend falls back to
+// Ollama, which is what every existing deployment runs.
+func New(cfg Config) Client {
+	if cfg.Timeout <= 0 {
+		cfg.Timeout = DefaultTimeout
+	}
+	base := backend{
+		endpoint: strings.TrimRight(cfg.Endpoint, "/"),
+		model:    cfg.Model,
+		timeout:  cfg.Timeout,
+	}
+	switch cfg.Backend {
+	case "vllm":
+		return &VLLMClient{base}
+	default:
+		return &OllamaClient{base}
+	}
+}
+
+// backend holds the fields shared by both concrete clients.
+type backend struct {
+	endpoint string
+	model    string
+	timeout  time.Duration
+}
+
+func (b backend) Model() string { return b.model }
+
+// timeoutFor lets a single call widen or narrow the client default. The two
+// dispatch-voice callers rely on this: a dispatch is authored on a game
+// chokepoint and must not stall it, while a run summary rides a background
+// ticker and can afford a bigger model.
+func (b backend) timeoutFor(req Request) time.Duration {
+	if req.Timeout > 0 {
+		return req.Timeout
+	}
+	return b.timeout
+}
+
+// StripThink removes a leading <think>...</think> reasoning block, which Qwen
+// models emit even when thinking is disabled by some backends. Callers that
+// parse JSON out of the completion depend on this running first.
+func StripThink(s string) string {
+	for {
+		i := strings.Index(s, "<think>")
+		if i < 0 {
+			break
+		}
+		j := strings.Index(s, "</think>")
+		if j < 0 || j < i {
+			break
+		}
+		s = s[:i] + s[j+len("</think>"):]
+	}
+	return strings.TrimSpace(s)
+}
+
+func firstNonEmpty(vals ...string) string {
+	for _, v := range vals {
+		if v = strings.TrimSpace(v); v != "" {
+			return v
+		}
+	}
+	return ""
+}

internal/llm/ollama.go

@@ -0,0 +1,103 @@
+package llm
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"fmt"
+	"io"
+	"net/http"
+)
+
+// OllamaClient talks to Ollama's native /api/generate endpoint.
+type OllamaClient struct {
+	backend
+}
+
+func (c *OllamaClient) Backend() string { return "ollama" }
+
+type ollamaOptions struct {
+	NumCtx      int     `json:"num_ctx,omitempty"`
+	NumPredict  int     `json:"num_predict,omitempty"`
+	Temperature float64 `json:"temperature,omitempty"`
+}
+
+type ollamaRequest struct {
+	Model   string        `json:"model"`
+	Prompt  string        `json:"prompt"`
+	System  string        `json:"system,omitempty"`
+	Stream  bool          `json:"stream"`
+	Think   bool          `json:"think"`
+	Options ollamaOptions `json:"options,omitempty"`
+}
+
+// Generate posts a single non-streaming generation and returns the completion.
+func (c *OllamaClient) Generate(ctx context.Context, req Request) (string, error) {
+	ctx, cancel := context.WithTimeout(ctx, c.timeoutFor(req))
+	defer cancel()
+
+	body, err := json.Marshal(ollamaRequest{
+		Model:  c.model,
+		Prompt: req.Prompt,
+		System: req.System,
+		Stream: false,
+		Think:  false,
+		Options: ollamaOptions{
+			NumCtx:      req.NumCtx,
+			NumPredict:  req.MaxTokens,
+			Temperature: req.Temperature,
+		},
+	})
+	if err != nil {
+		return "", fmt.Errorf("ollama: marshal payload: %w", err)
+	}
+
+	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost,
+		c.endpoint+"/api/generate", bytes.NewReader(body))
+	if err != nil {
+		return "", fmt.Errorf("ollama: build request: %w", err)
+	}
+	httpReq.Header.Set("Content-Type", "application/json")
+
+	resp, err := http.DefaultClient.Do(httpReq)
+	if err != nil {
+		return "", fmt.Errorf("ollama request: %w", err)
+	}
+	defer resp.Body.Close()
+
+	respBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return "", fmt.Errorf("ollama: read response: %w", err)
+	}
+	if resp.StatusCode != http.StatusOK {
+		return "", fmt.Errorf("ollama HTTP %d: %s", resp.StatusCode, string(respBody))
+	}
+
+	var result struct {
+		Response string `json:"response"`
+	}
+	if err := json.Unmarshal(respBody, &result); err != nil {
+		return "", fmt.Errorf("ollama: parse response: %w", err)
+	}
+	return StripThink(result.Response), nil
+}
+
+// Ping lists locally installed models via Ollama's native /api/tags.
+func (c *OllamaClient) Ping(ctx context.Context) ([]string, error) {
+	ctx, cancel := context.WithTimeout(ctx, pingTimeout)
+	defer cancel()
+
+	var out struct {
+		Models []struct {
+			Name string `json:"name"`
+		} `json:"models"`
+	}
+	if err := getJSON(ctx, c.endpoint+"/api/tags", &out); err != nil {
+		return nil, err
+	}
+	names := make([]string, 0, len(out.Models))
+	for _, m := range out.Models {
+		names = append(names, m.Name)
+	}
+	return names, nil
+}

internal/llm/ping.go

@@ -0,0 +1,29 @@
+package llm
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"net/http"
+	"time"
+)
+
+// pingTimeout keeps a liveness probe short — /botinfo renders synchronously and
+// a hung endpoint must not hold the reply.
+const pingTimeout = 5 * time.Second
+
+func getJSON(ctx context.Context, url string, out any) error {
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
+	if err != nil {
+		return err
+	}
+	resp, err := http.DefaultClient.Do(req)
+	if err != nil {
+		return err
+	}
+	defer resp.Body.Close()
+	if resp.StatusCode != http.StatusOK {
+		return fmt.Errorf("HTTP %d", resp.StatusCode)
+	}
+	return json.NewDecoder(resp.Body).Decode(out)
+}

internal/llm/vllm.go

@@ -0,0 +1,116 @@
+package llm
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"fmt"
+	"io"
+	"net/http"
+)
+
+// VLLMClient talks to an OpenAI-compatible /v1/chat/completions endpoint.
+type VLLMClient struct {
+	backend
+}
+
+func (c *VLLMClient) Backend() string { return "vllm" }
+
+type vllmMessage struct {
+	Role    string `json:"role"`
+	Content string `json:"content"`
+}
+
+type vllmRequest struct {
+	Model       string        `json:"model"`
+	Messages    []vllmMessage `json:"messages"`
+	MaxTokens   int           `json:"max_tokens,omitempty"`
+	Temperature float64       `json:"temperature,omitempty"`
+	Stream      bool          `json:"stream"`
+	// ChatTemplateKwargs is a vLLM extension to the OpenAI schema. It is how
+	// Qwen3-family reasoning is switched off; the Ollama backend spells the
+	// same intent as its native "think": false.
+	ChatTemplateKwargs map[string]any `json:"chat_template_kwargs,omitempty"`
+}
+
+// Generate posts a single non-streaming completion and returns the message
+// content. The raw prompt is sent as one user message so the server-side chat
+// template still wraps it.
+func (c *VLLMClient) Generate(ctx context.Context, req Request) (string, error) {
+	ctx, cancel := context.WithTimeout(ctx, c.timeoutFor(req))
+	defer cancel()
+
+	msgs := make([]vllmMessage, 0, 2)
+	if req.System != "" {
+		msgs = append(msgs, vllmMessage{Role: "system", Content: req.System})
+	}
+	msgs = append(msgs, vllmMessage{Role: "user", Content: req.Prompt})
+
+	body, err := json.Marshal(vllmRequest{
+		Model:              c.model,
+		Messages:           msgs,
+		MaxTokens:          req.MaxTokens,
+		Temperature:        req.Temperature,
+		Stream:             false,
+		ChatTemplateKwargs: map[string]any{"enable_thinking": false},
+	})
+	if err != nil {
+		return "", fmt.Errorf("vllm: marshal payload: %w", err)
+	}
+
+	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost,
+		c.endpoint+"/v1/chat/completions", bytes.NewReader(body))
+	if err != nil {
+		return "", fmt.Errorf("vllm: build request: %w", err)
+	}
+	httpReq.Header.Set("Content-Type", "application/json")
+
+	resp, err := http.DefaultClient.Do(httpReq)
+	if err != nil {
+		return "", fmt.Errorf("vllm request: %w", err)
+	}
+	defer resp.Body.Close()
+
+	respBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return "", fmt.Errorf("vllm: read response: %w", err)
+	}
+	if resp.StatusCode != http.StatusOK {
+		return "", fmt.Errorf("vllm HTTP %d: %s", resp.StatusCode, string(respBody))
+	}
+
+	var result struct {
+		Choices []struct {
+			Message struct {
+				Content string `json:"content"`
+			} `json:"message"`
+		} `json:"choices"`
+	}
+	if err := json.Unmarshal(respBody, &result); err != nil {
+		return "", fmt.Errorf("vllm: parse response: %w", err)
+	}
+	if len(result.Choices) == 0 {
+		return "", fmt.Errorf("vllm: empty choices in response")
+	}
+	return StripThink(result.Choices[0].Message.Content), nil
+}
+
+// Ping lists served models via the OpenAI-compatible /v1/models.
+func (c *VLLMClient) Ping(ctx context.Context) ([]string, error) {
+	ctx, cancel := context.WithTimeout(ctx, pingTimeout)
+	defer cancel()
+
+	var out struct {
+		Data []struct {
+			ID string `json:"id"`
+		} `json:"data"`
+	}
+	if err := getJSON(ctx, c.endpoint+"/v1/models", &out); err != nil {
+		return nil, err
+	}
+	names := make([]string, 0, len(out.Data))
+	for _, m := range out.Data {
+		names = append(names, m.ID)
+	}
+	return names, nil
+}

internal/plugin/adventure.go

@@ -1374,6 +1374,7 @@ func (p *AdventurePlugin) sendTreasureDiscoveryDM(userID id.UserID, char *Advent
 		"{treasure_name}": def.Name,
 		"{bonus_desc}":    def.InventoryDesc,
 		"{location}":      loc.Name,
+		"{location_mid}":  advLocationMidSentence(loc.Name),
 	})
 
 	p.SendDM(userID, text)
@@ -1400,8 +1401,9 @@ func (p *AdventurePlugin) announceTreasureToRoom(char *AdventureCharacter, def *
 	}
 	displayName, _ := loadDisplayName(char.UserID)
 	announce := advSubstituteFlavor(def.RoomAnnounce, map[string]string{
-		"{name}":     displayName,
-		"{location}": loc.Name,
+		"{name}":         displayName,
+		"{location}":     loc.Name,
+		"{location_mid}": advLocationMidSentence(loc.Name),
 	})
 	p.SendMessage(id.RoomID(gr), announce)
 }

internal/plugin/adventure_flavor_treasure.go

@@ -232,7 +232,7 @@ var TreasureDiscovery = map[int][]string{
 	// No exclamation marks. No enthusiasm. Just weight.
 	5: {
 		"The {treasure_name}.\n\n" +
-			"You found it in the Abyssal Maw. It found you in the Abyssal Maw. " +
+			"You found it in {location_mid}. It found you in {location_mid}. " +
 			"The distinction matters less at depth.\n" +
 			"BONUS: {bonus_desc}.\n\n" +
 			"It came with you. Some things don't come with you. This one did. " +
@@ -240,15 +240,15 @@ var TreasureDiscovery = map[int][]string{
 
 		"The {treasure_name} is in your inventory.\n\n" +
 			"BONUS: {bonus_desc}.\n\n" +
-			"This is a Tier 5 rare. The Abyssal Maw doesn't give these up. " +
-			"The Abyssal Maw gave this one up. " +
+			"This is a Tier 5 rare. Nothing in {location_mid} gives these up. " +
+			"Something in {location_mid} gave this one up. " +
 			"That's a sentence worth sitting with.",
 
 		"You have the {treasure_name}.\n\n" +
 			"BONUS: {bonus_desc}.\n\n" +
-			"It was in the Abyssal Maw. The things that were between you and it " +
+			"It was in {location_mid}. The things that were between you and it " +
 			"are not between anything and anything anymore. " +
-			"The Maw is noting this. So is the item.",
+			"The place is noting this. So is the item.",
 
 		"The {treasure_name}.\n\n" +
 			"It's warm. It was warm when you found it, which it shouldn't be, " +
@@ -262,14 +262,14 @@ var TreasureDiscovery = map[int][]string{
 			"The {treasure_name}. {bonus_desc}.\n\n" +
 			"You have it. " +
 			"Keep it somewhere it won't be lost. " +
-			"The Abyssal Maw does not give second chances.",
+			"There are no second chances in {location_mid}.",
 
 		"The {treasure_name} is yours.\n\n" +
 			"BONUS: {bonus_desc}.\n\n" +
-			"The Abyssal Maw had it. You have it now. " +
-			"The Maw is aware of the transfer. " +
-			"The Maw is considering its position. " +
-			"You should not be in the Maw when it finishes considering.",
+			"It sat in {location_mid} until you took it. You have it now. " +
+			"The place is aware of the transfer. " +
+			"The place is considering its position. " +
+			"You should not be in {location_mid} when it finishes considering.",
 	},
 }
 

internal/plugin/adventure_render.go

@@ -70,6 +70,18 @@ func advClearFlavorHistory() {
 	})
 }
 
+// advLocationMidSentence renders a location display name for use after a
+// preposition: "The Abyssal Maw" becomes "the Abyssal Maw" so a line reads
+// "found in the Abyssal Maw" rather than "found in The Abyssal Maw".
+// Templates opt in with {location_mid}; {location} keeps the display form for
+// sentence-initial and "The {location}" positions.
+func advLocationMidSentence(name string) string {
+	if strings.HasPrefix(name, "The ") {
+		return "the " + strings.TrimPrefix(name, "The ")
+	}
+	return name
+}
+
 // advSubstituteFlavor replaces {var} placeholders in a flavor text string.
 func advSubstituteFlavor(template string, vars map[string]string) string {
 	pairs := make([]string, 0, len(vars)*2)

internal/plugin/adventure_treasure.go

@@ -46,6 +46,16 @@ var advTreasureDropRates = map[int]float64{
 	3: 0.008,
 	4: 0.004,
 	5: 0.0015,
+	// Tier 6 (Mythic post-game) breaks the downward curve on purpose. The rate
+	// per roll falls tier over tier because the lower tiers are ground daily;
+	// a Mythic run is gated behind L18 and both T5 bosses and happens rarely,
+	// so the same declining rate would mean a postgame player effectively never
+	// sees a treasure. Slightly above T5 keeps the per-run odds in the band the
+	// other tiers land in.
+	//
+	// NOTE: this rate is inert until advAllTreasures gains a tier 6 pool — see
+	// the TODO there. A tier with a rate but no pool drops nothing.
+	6: 0.002,
 }
 
 const advMaxTreasures = 3
@@ -213,7 +223,7 @@ var advAllTreasures = map[int][]AdvTreasureDef{
 				{Type: "success_chance", Value: 10},
 			},
 			InventoryDesc: "[THUNDERFURY, BLESSED BLADE OF THE WINDSEEKER]. Yes you got it. +12 Combat.",
-			RoomAnnounce:  "⚡ Did {name} get Thunderfury? {name} got Thunderfury. [THUNDERFURY, BLESSED BLADE OF THE WINDSEEKER] has been found in {location}.",
+			RoomAnnounce:  "⚡ Did {name} get Thunderfury? {name} got Thunderfury. [THUNDERFURY, BLESSED BLADE OF THE WINDSEEKER] has been found in {location_mid}.",
 		},
 		{
 			Key: "ocarina", Name: "The Ocarina (Cracked, Still Plays)", Tier: 4,
@@ -221,6 +231,13 @@ var advAllTreasures = map[int][]AdvTreasureDef{
 			InventoryDesc: "The Ocarina (Cracked). Three songs. +10 all skills. Do not play the third one.",
 		},
 	},
+	// TODO(t6-treasures): there is no tier 6 pool yet, so Mythic zones drop no
+	// treasure at all — advTreasureDropRates has a tier 6 rate waiting for it.
+	// What's missing is the content: four Mythic treasures with bonuses above
+	// the T5 line, InventoryDesc + RoomAnnounce strings for each (RoomAnnounce
+	// must use {location_mid}, never a literal zone name), and a TIER 6 register
+	// in TreasureDiscovery. Write them against internal/flavor/VOICE_CANON.md;
+	// the T5 pool below is the tonal floor to clear, not the ceiling.
 	5: {
 		{
 			Key: "shard_of_unnamed", Name: "Shard of the Unnamed", Tier: 5,
@@ -230,13 +247,13 @@ var advAllTreasures = map[int][]AdvTreasureDef{
 				{Type: "death_chance", Value: -5},
 			},
 			InventoryDesc: "Shard of the Unnamed. +15 Combat, +10% XP, -5% death.",
-			RoomAnnounce:  "🔴 {name} has recovered the Shard of the Unnamed from the Abyssal Maw. The server feels different.",
+			RoomAnnounce:  "🔴 {name} has recovered the Shard of the Unnamed from {location_mid}. The server feels different.",
 		},
 		{
 			Key: "cartographers_final_map", Name: "The Cartographer's Final Map", Tier: 5,
 			Bonuses:       []advTreasureBonusDef{{Type: "all_skills", Value: 12}},
 			InventoryDesc: "The Cartographer's Final Map. Updates on its own. +12 all skills, full map.",
-			RoomAnnounce:  "🔴 {name} has found the Cartographer's Final Map in the Abyssal Maw. It has their name on it. It always did.",
+			RoomAnnounce:  "🔴 {name} has found the Cartographer's Final Map in {location_mid}. It has their name on it. It always did.",
 		},
 		{
 			Key: "triforce_shard", Name: "The Triforce Shard (One Third of Something Larger)", Tier: 5,
@@ -246,7 +263,7 @@ var advAllTreasures = map[int][]AdvTreasureDef{
 				// Note: +15 to chosen skill is v2 interactive
 			},
 			InventoryDesc: "Triforce Shard (×1/3). Warm. Waiting. +5 all skills, -8% death.",
-			RoomAnnounce:  "🔺 {name} has recovered a Triforce Shard from the Abyssal Maw. One third of something. The other two thirds are somewhere. Probably.",
+			RoomAnnounce:  "🔺 {name} has recovered a Triforce Shard from {location_mid}. One third of something. The other two thirds are somewhere. Probably.",
 		},
 		{
 			Key: "the_corridor", Name: "The Corridor (You Know the One)", Tier: 5,
@@ -255,7 +272,7 @@ var advAllTreasures = map[int][]AdvTreasureDef{
 				{Type: "special_monthly_death_bypass", Value: 1}, // v2
 			},
 			InventoryDesc: "The Corridor. Folded. Don't look back. +12 all skills, monthly death bypass.",
-			RoomAnnounce:  "🔴 {name} found The Corridor in the Abyssal Maw. They know the one. So does it.",
+			RoomAnnounce:  "🔴 {name} found The Corridor in {location_mid}. They know the one. So does it.",
 		},
 	},
 }
@@ -285,7 +302,12 @@ func rollAdvTreasureDropDetailed(tier int, userID id.UserID, chatLevel int, weig
 
 	pool, ok := advAllTreasures[tier]
 	if !ok || len(pool) == 0 {
-		return nil, roll, rate
+		// A tier that has a rate but no pool (tier 6, until the TODO above is
+		// filled) cannot drop anything. Report a zero rate rather than the
+		// configured one: the caller turns a close roll into a "just missed"
+		// DM, and telling a postgame player they nearly won a treasure that
+		// cannot be won is worse than staying quiet.
+		return nil, 0, 0
 	}
 
 	// Pick random treasure

internal/plugin/adventure_treasure_test.go

@@ -66,6 +66,36 @@ func TestAdvTreasureDropDetailed_ForcedRollGrantsTreasure(t *testing.T) {
 	}
 }
 
+// TestAdvTreasureTier6_PoolStillMissing pins the known tier 6 gap: Mythic zones
+// carry a drop rate but no treasure pool, so a forced roll yields nothing and
+// the player is told nothing either (a "just missed" DM for an unwinnable
+// treasure would be a lie).
+//
+// This test is the reminder. Writing advAllTreasures[6] — see TODO(t6-treasures)
+// in adventure_treasure.go — turns it red, and the fix is to delete it and
+// assert the tier 6 drop instead, the way ForcedRollGrantsTreasure does for
+// tier 1.
+func TestAdvTreasureTier6_PoolStillMissing(t *testing.T) {
+	if err := db.Init(t.TempDir()); err != nil {
+		t.Fatalf("db.Init: %v", err)
+	}
+	if _, ok := advAllTreasures[6]; ok {
+		t.Fatal("a tier 6 treasure pool now exists — drop this test and assert the drop instead")
+	}
+	// The rate is configured and waiting, so the gap is the pool alone.
+	if advTreasureDropRates[6] == 0 {
+		t.Error("tier 6 drop rate went missing; a Mythic pool would be unreachable")
+	}
+	drop, roll, rate := rollAdvTreasureDropDetailed(6, "@mythic:example.org", 0, 1000)
+	if drop != nil {
+		t.Fatalf("tier 6 produced a drop from an empty pool: %+v", drop.Def)
+	}
+	// Zeroed, not merely no-drop: this is what keeps the near-miss DM quiet.
+	if roll != 0 || rate != 0 {
+		t.Errorf("empty pool reported roll=%v rate=%v, want 0/0 so no near-miss fires", roll, rate)
+	}
+}
+
 // The treasure and masterwork systems predate zones and speak AdvLocation.
 func TestAdvLocForZone(t *testing.T) {
 	loc := advLocForZone(ZoneGoblinWarrens)

internal/plugin/botinfo.go

@@ -1,12 +1,9 @@
 package plugin
 
 import (
-	"encoding/json"
+	"context"
 	"fmt"
-	"io"
 	"log/slog"
-	"net/http"
-	"os"
 	"strings"
 	"sync/atomic"
 	"time"
@@ -121,10 +118,8 @@ func (p *BotInfoPlugin) handleBotInfo(ctx MessageContext) error {
 	sb.WriteString(fmt.Sprintf("Active reminders: %d\n", activeReminders))
 
 	// LLM status
-	ollamaHost := os.Getenv("OLLAMA_HOST")
-	if ollamaHost != "" {
-		llmStatus := p.checkLLMStatus(ollamaHost)
-		sb.WriteString(fmt.Sprintf("LLM status: %s\n", llmStatus))
+	if llmConfigured() {
+		sb.WriteString(fmt.Sprintf("LLM status: %s\n", p.checkLLMStatus()))
 	} else {
 		sb.WriteString("LLM status: not configured\n")
 	}
@@ -159,42 +154,16 @@ func (p *BotInfoPlugin) handleBotInfo(ctx MessageContext) error {
 	return p.SendReply(ctx.RoomID, ctx.EventID, sb.String())
 }
 
-func (p *BotInfoPlugin) checkLLMStatus(ollamaHost string) string {
-	client := &http.Client{Timeout: 5 * time.Second}
-	apiURL := strings.TrimRight(ollamaHost, "/") + "/api/tags"
-
-	resp, err := client.Get(apiURL)
+// checkLLMStatus reports backend liveness for /botinfo. The endpoint it probes
+// differs per backend, which the llm package hides behind Ping.
+func (p *BotInfoPlugin) checkLLMStatus() string {
+	c := llmClient()
+	models, err := c.Ping(context.Background())
 	if err != nil {
-		return fmt.Sprintf("offline (%s)", err.Error())
+		return fmt.Sprintf("offline (%s: %s)", c.Backend(), err.Error())
 	}
-	defer resp.Body.Close()
-
-	if resp.StatusCode != 200 {
-		return fmt.Sprintf("error (HTTP %d)", resp.StatusCode)
+	if len(models) == 0 {
+		return fmt.Sprintf("online (%s, no models loaded)", c.Backend())
 	}
-
-	body, err := io.ReadAll(resp.Body)
-	if err != nil {
-		return "online (could not read response)"
-	}
-
-	var result struct {
-		Models []struct {
-			Name string `json:"name"`
-		} `json:"models"`
-	}
-	if err := json.Unmarshal(body, &result); err != nil {
-		return "online (could not parse response)"
-	}
-
-	modelNames := make([]string, 0, len(result.Models))
-	for _, m := range result.Models {
-		modelNames = append(modelNames, m.Name)
-	}
-
-	if len(modelNames) == 0 {
-		return "online (no models loaded)"
-	}
-
-	return fmt.Sprintf("online (%d models: %s)", len(modelNames), strings.Join(modelNames, ", "))
+	return fmt.Sprintf("online (%s, %d models: %s)", c.Backend(), len(models), strings.Join(models, ", "))
 }

internal/plugin/hangman.go

@@ -856,9 +856,7 @@ func (p *HangmanPlugin) handleSubmit(ctx MessageContext, phrase string) error {
 	}
 
 	// LLM screening
-	ollamaHost := os.Getenv("OLLAMA_HOST")
-	ollamaModel := os.Getenv("OLLAMA_MODEL")
-	if ollamaHost == "" || ollamaModel == "" {
+	if !llmConfigured() {
 		// No LLM available — add directly
 		if err := p.addPhrase(phrase); err != nil {
 			if err.Error() == "duplicate phrase" {
@@ -879,7 +877,7 @@ or
 
 Phrase: %s`, phrase)
 
-	result, err := callOllama(ollamaHost, ollamaModel, prompt)
+	result, err := callLLM(prompt)
 	if err != nil {
 		slog.Error("hangman: LLM screening failed", "err", err)
 		// Fail open — add it

internal/plugin/holdem_tips.go

@@ -1,24 +1,24 @@
 package plugin
 
 import (
-	"bytes"
-	"encoding/json"
+	"context"
 	"fmt"
-	"io"
 	"log/slog"
-	"net/http"
-	"os"
 	"regexp"
 	"strings"
 	"time"
 
 	"gogobee/internal/db"
+	"gogobee/internal/llm"
 
 	"github.com/chehsunliu/poker"
 	"maunium.net/go/mautrix/id"
 )
 
-var holdemTipsClient = &http.Client{Timeout: 60 * time.Second}
+// holdemTipTimeout preserves the 60s budget the tip rewriter's own http.Client
+// enforced. Tips are delivered as private messages during a hand, so this sits
+// between the passive 30s paths and the interactive 120s default.
+const holdemTipTimeout = 60 * time.Second
 
 // loadTipsPref loads a user's tip preference from the database.
 func loadTipsPref(userID id.UserID) bool {
@@ -491,11 +491,8 @@ func cardSuitIndex(c poker.Card) int {
 func generateTip(ctx holdemTipContext) string {
 	base := generateRulesTip(ctx)
 
-	host := os.Getenv("OLLAMA_HOST")
-	model := os.Getenv("OLLAMA_MODEL")
-
-	if host != "" && model != "" {
-		rewritten, err := rewriteTipWithLLM(host, model, ctx, base)
+	if llmConfigured() {
+		rewritten, err := rewriteTipWithLLM(ctx, base)
 		if err != nil {
 			slog.Warn("holdem: LLM tip rewrite failed, using rules tip", "err", err)
 		} else if rewritten != "" {
@@ -626,40 +623,19 @@ func buildTipUserPrompt(ctx holdemTipContext) string {
 // variety. The rules tip is the source of truth — if the rewrite diverges
 // (empty, action vocabulary changed, etc.) we reject it and the caller falls
 // back to the original.
-func rewriteTipWithLLM(host, model string, ctx holdemTipContext, base string) (string, error) {
+func rewriteTipWithLLM(ctx holdemTipContext, base string) (string, error) {
 	userMsg := buildTipUserPrompt(ctx) + "\nTIP:\n" + base + "\n"
-	req := ollamaChatRequest{
-		Model: model,
-		Messages: []chatMessage{
-			{Role: "system", Content: buildTipSystemPrompt()},
-			{Role: "user", Content: userMsg},
-		},
-		Stream: false,
-	}
 
-	body, err := json.Marshal(req)
+	raw, err := llmGenerate(context.Background(), llm.Request{
+		System:  buildTipSystemPrompt(),
+		Prompt:  userMsg,
+		Timeout: holdemTipTimeout,
+	})
 	if err != nil {
-		return "", fmt.Errorf("marshal: %w", err)
+		return "", err
 	}
 
-	url := strings.TrimRight(host, "/") + "/api/chat"
-	resp, err := holdemTipsClient.Post(url, "application/json", bytes.NewReader(body))
-	if err != nil {
-		return "", fmt.Errorf("request: %w", err)
-	}
-	defer resp.Body.Close()
-
-	if resp.StatusCode != http.StatusOK {
-		respBody, _ := io.ReadAll(resp.Body)
-		return "", fmt.Errorf("status %d: %s", resp.StatusCode, string(respBody))
-	}
-
-	var ollamaResp ollamaChatResponse
-	if err := json.NewDecoder(resp.Body).Decode(&ollamaResp); err != nil {
-		return "", fmt.Errorf("decode: %w", err)
-	}
-
-	tip := extractTipFromResponse(ollamaResp.Message.Content)
+	tip := extractTipFromResponse(raw)
 	if tip == "" {
 		return "", fmt.Errorf("empty response")
 	}

internal/plugin/howami.go

@@ -1,18 +1,15 @@
 package plugin
 
 import (
-	"bytes"
+	"context"
 	"database/sql"
-	"encoding/json"
 	"fmt"
-	"io"
 	"log/slog"
-	"net/http"
 	"os"
 	"strings"
-	"time"
 
 	"gogobee/internal/db"
+	"gogobee/internal/llm"
 
 	"maunium.net/go/mautrix"
 	"maunium.net/go/mautrix/id"
@@ -47,9 +44,7 @@ func (p *HowAmIPlugin) OnMessage(ctx MessageContext) error {
 		return nil
 	}
 
-	ollamaHost := os.Getenv("OLLAMA_HOST")
-	ollamaModel := os.Getenv("OLLAMA_MODEL")
-	if ollamaHost == "" || ollamaModel == "" {
+	if !llmConfigured() {
 		return p.SendReply(ctx.RoomID, ctx.EventID, "LLM is not configured.")
 	}
 
@@ -84,9 +79,9 @@ Write the roast now. Do not include any preamble or explanation, just the roast
 			botName, string(target), profile,
 		)
 
-		response, err := callOllama(ollamaHost, ollamaModel, prompt)
+		response, err := callLLM(prompt)
 		if err != nil {
-			slog.Error("howami: ollama call", "err", err)
+			slog.Error("howami: llm call", "err", err)
 			p.SendReply(ctx.RoomID, ctx.EventID, "Couldn't generate the profile. Thanks, Ollama.")
 			return
 		}
@@ -181,55 +176,13 @@ func (p *HowAmIPlugin) gatherProfile(userID id.UserID) string {
 	return sb.String()
 }
 
-// callOllama sends a prompt to the Ollama generate endpoint and returns the response.
-func callOllama(host, model, prompt string) (string, error) {
-	apiURL := strings.TrimRight(host, "/") + "/api/generate"
-
-	payload := map[string]interface{}{
-		"model":  model,
-		"prompt": prompt,
-		"stream": false,
-		"think":  false,
-		"options": map[string]interface{}{
-			"num_ctx": 8192,
-		},
-	}
-
-	data, err := json.Marshal(payload)
-	if err != nil {
-		return "", fmt.Errorf("marshal payload: %w", err)
-	}
-
-	client := &http.Client{Timeout: 120 * time.Second}
-	resp, err := client.Post(apiURL, "application/json", bytes.NewReader(data))
-	if err != nil {
-		return "", fmt.Errorf("ollama request: %w", err)
-	}
-	defer resp.Body.Close()
-
-	body, err := io.ReadAll(resp.Body)
-	if err != nil {
-		return "", fmt.Errorf("read response: %w", err)
-	}
-
-	if resp.StatusCode != 200 {
-		return "", fmt.Errorf("ollama HTTP %d: %s", resp.StatusCode, string(body))
-	}
-
-	var result struct {
-		Response string `json:"response"`
-	}
-	if err := json.Unmarshal(body, &result); err != nil {
-		return "", fmt.Errorf("parse response: %w", err)
-	}
-
-	response := result.Response
-	// Strip <think>...</think> blocks (Qwen 3.5 reasoning)
-	if i := strings.Index(response, "<think>"); i != -1 {
-		if j := strings.Index(response, "</think>"); j != -1 {
-			response = response[:i] + response[j+len("</think>"):]
-		}
-	}
-
-	return strings.TrimSpace(response), nil
+// callLLM sends a prompt to whichever backend is configured and returns the
+// completion. Reasoning blocks are stripped by the client. The 8192 context
+// hint is what this path has always asked Ollama for; vLLM ignores it and uses
+// the window fixed at server launch.
+func callLLM(prompt string) (string, error) {
+	return llmGenerate(context.Background(), llm.Request{
+		Prompt: prompt,
+		NumCtx: 8192,
+	})
 }

internal/plugin/llm_client.go

@@ -0,0 +1,45 @@
+package plugin
+
+import (
+	"context"
+	"sync"
+
+	"gogobee/internal/llm"
+)
+
+// The inference backend is process-wide: one endpoint, one model, selected by
+// env at startup. Plugins share a single client rather than each rebuilding one
+// per invocation, and read config through llmConfigured/llmGenerate so that
+// swapping Ollama for vLLM is a config change rather than a code change.
+var (
+	llmOnce   sync.Once
+	llmShared llm.Client
+	llmCfg    llm.Config
+)
+
+func llmInit() {
+	llmOnce.Do(func() {
+		llmCfg = llm.ConfigFromEnv()
+		llmShared = llm.New(llmCfg)
+	})
+}
+
+// llmConfigured reports whether an endpoint and model are set. Plugins call
+// this to stay dormant instead of erroring on every invocation — the same role
+// the old `if ollamaHost == "" || ollamaModel == ""` guards played.
+func llmConfigured() bool {
+	llmInit()
+	return llmCfg.Configured()
+}
+
+// llmClient returns the shared backend client.
+func llmClient() llm.Client {
+	llmInit()
+	return llmShared
+}
+
+// llmGenerate is the one-line path for the common case: a raw prompt in,
+// visible completion out, reasoning blocks already stripped.
+func llmGenerate(ctx context.Context, req llm.Request) (string, error) {
+	return llmClient().Generate(ctx, req)
+}

internal/plugin/llm_passive.go

@@ -1,13 +1,11 @@
 package plugin
 
 import (
-	"bytes"
+	"context"
 	"encoding/json"
 	"fmt"
-	"io"
 	"log/slog"
 	"math/rand"
-	"net/http"
 	"os"
 	"regexp"
 	"strconv"
@@ -18,6 +16,7 @@ import (
 
 	"gogobee/internal/db"
 	"gogobee/internal/dreamclient"
+	"gogobee/internal/llm"
 
 	"maunium.net/go/mautrix"
 	"maunium.net/go/mautrix/id"
@@ -54,29 +53,31 @@ type queueItem struct {
 	FormattedBody string
 }
 
-// LLMPassivePlugin classifies messages using Ollama and reacts accordingly.
+// classifyTimeout is the per-message budget for passive classification. It
+// preserves the 30s cap the plugin's own http.Client used to enforce, which is
+// deliberately tighter than the interactive default: classification runs on
+// sampled traffic and must never back up the queue.
+const classifyTimeout = 30 * time.Second
+
+// LLMPassivePlugin classifies messages using the configured LLM backend and
+// reacts accordingly.
 type LLMPassivePlugin struct {
 	Base
-	xp          *XPPlugin
-	dict        *dreamclient.Client
-	ollamaHost  string
-	ollamaModel string
-	sampleRate  float64
-	enabled     bool
+	xp         *XPPlugin
+	dict       *dreamclient.Client
+	sampleRate float64
+	enabled    bool
 
 	mu      sync.Mutex
 	queue   []queueItem
 	backoff time.Duration
 
-	httpClient *http.Client
-	stopCh     chan struct{}
+	stopCh chan struct{}
 }
 
 // NewLLMPassivePlugin creates a new LLM passive classification plugin.
 func NewLLMPassivePlugin(client *mautrix.Client, xp *XPPlugin, dict *dreamclient.Client) *LLMPassivePlugin {
-	host := os.Getenv("OLLAMA_HOST")
-	model := os.Getenv("OLLAMA_MODEL")
-	enabled := host != "" && model != ""
+	enabled := llmConfigured()
 
 	sampleRate := 0.15
 	if v := os.Getenv("LLM_SAMPLE_RATE"); v != "" {
@@ -86,16 +87,13 @@ func NewLLMPassivePlugin(client *mautrix.Client, xp *XPPlugin, dict *dreamclient
 	}
 
 	p := &LLMPassivePlugin{
-		Base:        NewBase(client),
-		xp:          xp,
-		dict:        dict,
-		ollamaHost:  host,
-		ollamaModel: model,
-		sampleRate:  sampleRate,
-		enabled:     enabled,
-		backoff:     5 * time.Second,
-		httpClient:  &http.Client{Timeout: 30 * time.Second},
-		stopCh:      make(chan struct{}),
+		Base:       NewBase(client),
+		xp:         xp,
+		dict:       dict,
+		sampleRate: sampleRate,
+		enabled:    enabled,
+		backoff:    5 * time.Second,
+		stopCh:     make(chan struct{}),
 	}
 
 	return p
@@ -116,11 +114,11 @@ func (p *LLMPassivePlugin) Commands() []CommandDef {
 
 func (p *LLMPassivePlugin) Init() error {
 	if p.enabled {
-		slog.Info("llm_passive: enabled", "host", p.ollamaHost, "model", p.ollamaModel, "sample_rate", p.sampleRate)
+		slog.Info("llm_passive: enabled", "backend", llmClient().Backend(),
+			"model", llmClient().Model(), "sample_rate", p.sampleRate)
 		go p.processQueue()
 	} else {
-		slog.Warn("llm_passive: disabled (OLLAMA_HOST or OLLAMA_MODEL not set)",
-			"host", p.ollamaHost, "model", p.ollamaModel)
+		slog.Warn("llm_passive: disabled (LLM endpoint or model not set)")
 	}
 	return nil
 }
@@ -329,9 +327,9 @@ func (p *LLMPassivePlugin) classifyAndProcess(item queueItem) error {
 	var todayWOTD string
 	db.Get().QueryRow(`SELECT word FROM wotd_log WHERE date = ?`, today).Scan(&todayWOTD)
 
-	result, err := p.callOllama(item.Body+mentionHint, todayWOTD)
+	result, err := p.classify(item.Body+mentionHint, todayWOTD)
 	if err != nil {
-		return fmt.Errorf("ollama call: %w", err)
+		return fmt.Errorf("llm call: %w", err)
 	}
 
 	// Resolve any display names in LLM targets back to MXIDs
@@ -464,21 +462,9 @@ func (p *LLMPassivePlugin) classifyAndProcess(item queueItem) error {
 	return nil
 }
 
-// ollamaRequest is the request body for the Ollama API.
-type ollamaRequest struct {
-	Model  string `json:"model"`
-	Prompt string `json:"prompt"`
-	Stream bool   `json:"stream"`
-	Think  bool   `json:"think"`
-}
-
-// ollamaResponse is the response from the Ollama API.
-type ollamaResponse struct {
-	Response string `json:"response"`
-}
-
-// callOllama sends a classification prompt to Ollama and parses the JSON result.
-func (p *LLMPassivePlugin) callOllama(messageText, wotd string) (*classificationResult, error) {
+// classify sends a classification prompt to the configured backend and parses
+// the JSON result.
+func (p *LLMPassivePlugin) classify(messageText, wotd string) (*classificationResult, error) {
 	wotdInstruction := `"wotd_used": false`
 	if wotd != "" {
 		wotdInstruction = fmt.Sprintf(`"wotd_used": true | false (whether the message uses the word "%s" correctly and meaningfully — not just mentioning or quoting it)`, wotd)
@@ -500,36 +486,18 @@ JSON schema:
 
 Message: %s`, wotdInstruction, messageText)
 
-	reqBody := ollamaRequest{
-		Model:  p.ollamaModel,
-		Prompt: prompt,
-		Stream: false,
-	}
-
-	body, err := json.Marshal(reqBody)
+	slog.Debug("llm_passive: calling backend", "backend", llmClient().Backend(), "model", llmClient().Model())
+	// Classification rides the passive path on every sampled message, so it keeps
+	// the tighter budget it always had rather than the interactive default.
+	raw, err := llmGenerate(context.Background(), llm.Request{
+		Prompt:  prompt,
+		Timeout: classifyTimeout,
+	})
 	if err != nil {
-		return nil, fmt.Errorf("marshal request: %w", err)
+		return nil, err
 	}
 
-	url := strings.TrimRight(p.ollamaHost, "/") + "/api/generate"
-	slog.Debug("llm_passive: calling ollama", "url", url, "model", p.ollamaModel)
-	resp, err := p.httpClient.Post(url, "application/json", bytes.NewReader(body))
-	if err != nil {
-		return nil, fmt.Errorf("ollama request: %w", err)
-	}
-	defer resp.Body.Close()
-
-	if resp.StatusCode != http.StatusOK {
-		respBody, _ := io.ReadAll(resp.Body)
-		return nil, fmt.Errorf("ollama status %d: %s", resp.StatusCode, string(respBody))
-	}
-
-	var ollamaResp ollamaResponse
-	if err := json.NewDecoder(resp.Body).Decode(&ollamaResp); err != nil {
-		return nil, fmt.Errorf("decode ollama response: %w", err)
-	}
-
-	result, err := parseClassification(ollamaResp.Response)
+	result, err := parseClassification(raw)
 	if err != nil {
 		return nil, fmt.Errorf("parse classification: %w", err)
 	}

internal/plugin/market.go

@@ -1,10 +1,9 @@
 package plugin
 
 import (
-	"bytes"
+	"context"
 	"encoding/json"
 	"fmt"
-	"io"
 	"log/slog"
 	"math"
 	"net/http"
@@ -15,6 +14,7 @@ import (
 	"time"
 
 	"gogobee/internal/db"
+	"gogobee/internal/llm"
 
 	"maunium.net/go/mautrix"
 	"maunium.net/go/mautrix/id"
@@ -311,9 +311,7 @@ Do not use em dashes. Do not use exclamation marks. Do not offer financial advic
 If markets are closed or data is stale, note it briefly and move on.`
 
 func (p *MarketPlugin) generateDailySummary(date string) string {
-	host := os.Getenv("OLLAMA_HOST")
-	model := os.Getenv("OLLAMA_MODEL")
-	if host == "" || model == "" {
+	if !llmConfigured() {
 		return ""
 	}
 
@@ -349,7 +347,7 @@ func (p *MarketPlugin) generateDailySummary(date string) string {
 	}
 	prompt.WriteString("\nWrite a 2-3 sentence summary.")
 
-	result, err := p.callOllamaChat(host, model, marketSystemPrompt, prompt.String())
+	result, err := p.chatLLM(marketSystemPrompt, prompt.String())
 	if err != nil {
 		slog.Error("market: ollama summary failed", "err", err)
 		return ""
@@ -358,9 +356,7 @@ func (p *MarketPlugin) generateDailySummary(date string) string {
 }
 
 func (p *MarketPlugin) generateReportSummary(snapsByDate map[string][]marketSnapshot, dateRange string) string {
-	host := os.Getenv("OLLAMA_HOST")
-	model := os.Getenv("OLLAMA_MODEL")
-	if host == "" || model == "" {
+	if !llmConfigured() {
 		return ""
 	}
 
@@ -395,7 +391,7 @@ func (p *MarketPlugin) generateReportSummary(snapsByDate map[string][]marketSnap
 	}
 	prompt.WriteString("\nDescribe the trend in 2-3 sentences. Note any significant moves or divergences between indices.\nBe sardonic but accurate. Reference the VIX trajectory when relevant.")
 
-	result, err := p.callOllamaChat(host, model, marketSystemPrompt, prompt.String())
+	result, err := p.chatLLM(marketSystemPrompt, prompt.String())
 	if err != nil {
 		slog.Warn("market: ollama report summary failed", "err", err)
 		return ""
@@ -403,49 +399,14 @@ func (p *MarketPlugin) generateReportSummary(snapsByDate map[string][]marketSnap
 	return result
 }
 
-// callOllamaChat calls the Ollama /api/chat endpoint with a system and user message.
-// Uses the types already defined in holdem_tips.go (same package).
-func (p *MarketPlugin) callOllamaChat(host, model, systemPrompt, userPrompt string) (string, error) {
-	req := ollamaChatRequest{
-		Model: model,
-		Messages: []chatMessage{
-			{Role: "system", Content: systemPrompt},
-			{Role: "user", Content: userPrompt},
-		},
-		Stream: false,
-	}
-
-	body, err := json.Marshal(req)
-	if err != nil {
-		return "", fmt.Errorf("marshal: %w", err)
-	}
-
-	url := strings.TrimRight(host, "/") + "/api/chat"
-	resp, err := p.httpClient.Post(url, "application/json", bytes.NewReader(body))
-	if err != nil {
-		return "", fmt.Errorf("request: %w", err)
-	}
-	defer resp.Body.Close()
-
-	if resp.StatusCode != http.StatusOK {
-		respBody, _ := io.ReadAll(resp.Body)
-		return "", fmt.Errorf("status %d: %s", resp.StatusCode, string(respBody))
-	}
-
-	var ollamaResp ollamaChatResponse
-	if err := json.NewDecoder(resp.Body).Decode(&ollamaResp); err != nil {
-		return "", fmt.Errorf("decode: %w", err)
-	}
-
-	text := ollamaResp.Message.Content
-	// Strip <think>...</think> blocks (reasoning models)
-	if i := strings.Index(text, "<think>"); i != -1 {
-		if j := strings.Index(text, "</think>"); j != -1 {
-			text = text[:i] + text[j+len("</think>"):]
-		}
-	}
-
-	return strings.TrimSpace(text), nil
+// chatLLM sends a system+user pair to the configured backend. Both backends
+// map this onto their own chat shape, so the market summaries read the same
+// whichever one is serving.
+func (p *MarketPlugin) chatLLM(systemPrompt, userPrompt string) (string, error) {
+	return llmGenerate(context.Background(), llm.Request{
+		System: systemPrompt,
+		Prompt: userPrompt,
+	})
 }
 
 // ── DB Helpers ───────────────────────────────────────────────────────────────
@@ -916,12 +877,10 @@ func (p *MarketPlugin) handleVixReport(ctx MessageContext) error {
 	}
 
 	var summary string
-	host := os.Getenv("OLLAMA_HOST")
-	model := os.Getenv("OLLAMA_MODEL")
-	if host != "" && model != "" {
+	if llmConfigured() {
 		prompt := fmt.Sprintf("VIX (fear index) data over %d days (%s to %s):\n%s\n\nDescribe the fear/greed trajectory in 2-3 sentences. Be sardonic but accurate.",
 			len(entries), entries[0].Date, entries[len(entries)-1].Date, strings.Join(prices, ", "))
-		summary, _ = p.callOllamaChat(host, model, marketSystemPrompt, prompt)
+		summary, _ = p.chatLLM(marketSystemPrompt, prompt)
 	}
 
 	var sb strings.Builder
@@ -1020,12 +979,10 @@ func (p *MarketPlugin) handleCompare(ctx MessageContext, args []string) error {
 	for _, e := range entries {
 		prices = append(prices, fmt.Sprintf("%.2f", e.Price))
 	}
-	host := os.Getenv("OLLAMA_HOST")
-	model := os.Getenv("OLLAMA_MODEL")
-	if host != "" && model != "" {
+	if llmConfigured() {
 		prompt := fmt.Sprintf("%s over %d days (%s to %s):\n%s\n\nDescribe the trend in 2-3 sentences. Be sardonic but accurate.",
 			idx.DisplayName, len(entries), entries[0].Date, entries[len(entries)-1].Date, strings.Join(prices, ", "))
-		if summary, err := p.callOllamaChat(host, model, marketSystemPrompt, prompt); err == nil && summary != "" {
+		if summary, err := p.chatLLM(marketSystemPrompt, prompt); err == nil && summary != "" {
 			sb.WriteString(summary)
 			sb.WriteString("\n\n")
 		}

internal/plugin/pete_dispatch_voice.go

@@ -1,16 +1,14 @@
 package plugin
 
 import (
-	"bytes"
+	"context"
 	"encoding/json"
 	"fmt"
-	"io"
 	"log/slog"
-	"net/http"
-	"os"
 	"strings"
 	"time"
 
+	"gogobee/internal/llm"
 	"gogobee/internal/peteclient"
 )
 
@@ -39,22 +37,18 @@ const (
 	maxDispatchLede     = 800
 )
 
-var dispatchHTTP = &http.Client{Timeout: dispatchLLMTimeout}
-
 // authorDispatch turns a fact into a headline+lede in Pete's voice, or returns
 // two empty strings if the model is unconfigured, errors, times out, or produces
 // anything malformed. The fact must already have its FINAL Actors set (post
 // opt-out anonymisation) — that list is the only set of names the prose may use,
 // and it is what Pete's guard checks the output against.
 func authorDispatch(f peteclient.Fact) (headline, lede string) {
-	host := os.Getenv("OLLAMA_HOST")
-	model := os.Getenv("OLLAMA_MODEL")
-	if host == "" || model == "" {
+	if !llmConfigured() {
 		return "", ""
 	}
 
 	prompt := buildDispatchPrompt(f)
-	raw, err := callOllamaDispatch(dispatchHTTP, host, model, prompt)
+	raw, err := callLLMDispatch(dispatchLLMTimeout, prompt)
 	if err != nil {
 		slog.Warn("pete dispatch: LLM authoring failed, Pete will template", "guid", f.GUID, "err", err)
 		return "", ""
@@ -128,45 +122,17 @@ The event:
 %s`, names, facts.String())
 }
 
-// callOllamaDispatch posts a single non-streaming generation and returns the raw
-// completion (think-tags stripped). The client is a parameter because the two
-// callers have genuinely different patience: a dispatch is authored on a game
-// chokepoint and must not stall it, while a run summary rides a background
-// ticker and can afford to wait for a bigger model. See runSummaryHTTP.
-func callOllamaDispatch(client *http.Client, host, model, prompt string) (string, error) {
-	payload := map[string]interface{}{
-		"model":  model,
-		"prompt": prompt,
-		"stream": false,
-		"think":  false,
-		"options": map[string]interface{}{
-			"num_ctx": 4096,
-		},
-	}
-	data, err := json.Marshal(payload)
-	if err != nil {
-		return "", fmt.Errorf("marshal payload: %w", err)
-	}
-	apiURL := strings.TrimRight(host, "/") + "/api/generate"
-	resp, err := client.Post(apiURL, "application/json", bytes.NewReader(data))
-	if err != nil {
-		return "", fmt.Errorf("ollama request: %w", err)
-	}
-	defer resp.Body.Close()
-	body, err := io.ReadAll(resp.Body)
-	if err != nil {
-		return "", fmt.Errorf("read response: %w", err)
-	}
-	if resp.StatusCode != http.StatusOK {
-		return "", fmt.Errorf("ollama HTTP %d: %s", resp.StatusCode, string(body))
-	}
-	var result struct {
-		Response string `json:"response"`
-	}
-	if err := json.Unmarshal(body, &result); err != nil {
-		return "", fmt.Errorf("parse response: %w", err)
-	}
-	return result.Response, nil
+// callLLMDispatch posts a single non-streaming generation and returns the raw
+// completion. The timeout is a parameter because the two callers have genuinely
+// different patience: a dispatch is authored on a game chokepoint and must not
+// stall it, while a run summary rides a background ticker and can afford to wait
+// for a bigger model. See runSummaryTimeout.
+func callLLMDispatch(timeout time.Duration, prompt string) (string, error) {
+	return llmGenerate(context.Background(), llm.Request{
+		Prompt:  prompt,
+		NumCtx:  4096,
+		Timeout: timeout,
+	})
 }
 
 // parseDispatch pulls {headline, lede} out of the model's completion, tolerating

internal/plugin/pete_run_summary.go

@@ -4,8 +4,6 @@ import (
 	"encoding/json"
 	"fmt"
 	"log/slog"
-	"net/http"
-	"os"
 	"sort"
 	"strings"
 	"sync/atomic"
@@ -70,8 +68,6 @@ const maxRunSummary = 1200
 // for load-then-generate, and a timeout here really does mean the box is down.
 const runSummaryTimeout = 5 * time.Minute
 
-var runSummaryHTTP = &http.Client{Timeout: runSummaryTimeout}
-
 // runSummaryBusy is the whole concurrency story: at most one sweep in flight,
 // ever. The ticker starts one and moves on, so a cold model loading for minutes
 // costs the board nothing, and the ticks that fire meanwhile find the flag set
@@ -105,7 +101,7 @@ func (p *AdventurePlugin) sweepRunSummaries() {
 	if !peteclient.Enabled() || !newsEmissionOn() {
 		return
 	}
-	if os.Getenv("OLLAMA_HOST") == "" || os.Getenv("OLLAMA_MODEL") == "" {
+	if !llmConfigured() {
 		return // no model, no summary, no wasted queries asking which run needs one
 	}
 	runID := nextRunNeedingSummary()
@@ -174,8 +170,7 @@ func authorRunSummary(runID string) (summary, name string) {
 		return "", ""
 	}
 
-	raw, err := callOllamaDispatch(runSummaryHTTP, os.Getenv("OLLAMA_HOST"), os.Getenv("OLLAMA_MODEL"),
-		buildRunSummaryPrompt(name, beats))
+	raw, err := callLLMDispatch(runSummaryTimeout, buildRunSummaryPrompt(name, beats))
 	if err != nil {
 		slog.Warn("run summary: LLM authoring failed", "run", runID, "err", err)
 		return "", name

internal/plugin/plugin.go

@@ -83,9 +83,17 @@ func PluginVersion(p Plugin) string {
 }
 
 // dmCache maps user IDs to their DM room IDs to avoid creating duplicate rooms.
+// It fronts the dm_rooms table; notDMCache is the negative half, marking rooms
+// LearnDMRoom has already ruled out so group rooms cost one member lookup ever.
 var (
-	dmCache   = make(map[id.UserID]id.RoomID)
-	dmCacheMu sync.Mutex
+	dmCache    = make(map[id.UserID]id.RoomID)
+	dmMapped   = make(map[id.UserID]bool)
+	notDMCache = make(map[id.RoomID]bool)
+	dmCacheMu  sync.Mutex
+
+	// One-shot index of two-person rooms, built by findExistingDMRoom.
+	dmSweepOnce  sync.Once
+	dmSweepIndex = make(map[id.UserID]id.RoomID)
 )
 
 // Base provides common helpers for plugin implementations.
@@ -686,6 +694,180 @@ func (b *Base) SendReact(roomID id.RoomID, eventID id.EventID, emoji string) err
 	return err
 }
 
+// rememberDMRoom pins a user's DM room in both the in-process cache and the
+// database, so a restart doesn't send the bot off creating a duplicate room.
+func rememberDMRoom(userID id.UserID, roomID id.RoomID) {
+	dmCacheMu.Lock()
+	dmCache[userID] = roomID
+	dmMapped[userID] = true
+	dmCacheMu.Unlock()
+
+	if d := db.Get(); d != nil {
+		_, err := d.Exec(`INSERT INTO dm_rooms (user_id, room_id, updated_at)
+			VALUES (?, ?, unixepoch())
+			ON CONFLICT(user_id) DO UPDATE SET room_id = excluded.room_id, updated_at = excluded.updated_at`,
+			string(userID), string(roomID))
+		if err != nil {
+			slog.Error("persist dm room", "user", userID, "room", roomID, "err", err)
+		}
+	}
+}
+
+// forgetDMRoom drops a mapping that no longer resolves to a live shared room.
+func forgetDMRoom(userID id.UserID) {
+	dmCacheMu.Lock()
+	delete(dmCache, userID)
+	delete(dmMapped, userID)
+	dmCacheMu.Unlock()
+
+	if d := db.Get(); d != nil {
+		if _, err := d.Exec(`DELETE FROM dm_rooms WHERE user_id = ?`, string(userID)); err != nil {
+			slog.Error("forget dm room", "user", userID, "err", err)
+		}
+	}
+}
+
+// storedDMRoom reads the persisted DM room for a user, if any.
+func storedDMRoom(userID id.UserID) (id.RoomID, bool) {
+	d := db.Get()
+	if d == nil {
+		return "", false
+	}
+	var roomID string
+	err := d.QueryRow(`SELECT room_id FROM dm_rooms WHERE user_id = ?`, string(userID)).Scan(&roomID)
+	if err != nil || roomID == "" {
+		return "", false
+	}
+	return id.RoomID(roomID), true
+}
+
+// dmRoomUsable reports whether the bot and the user still share the room. The
+// user counts as present while merely invited — they often never accept, and
+// treating that as "gone" is what would recreate the room on every send.
+func (b *Base) dmRoomUsable(roomID id.RoomID, userID id.UserID) bool {
+	ctx := context.Background()
+
+	var self event.MemberEventContent
+	if err := b.Client.StateEvent(ctx, roomID, event.StateMember, string(b.Client.UserID), &self); err != nil {
+		return false
+	}
+	if self.Membership != event.MembershipJoin {
+		return false
+	}
+
+	var other event.MemberEventContent
+	if err := b.Client.StateEvent(ctx, roomID, event.StateMember, string(userID), &other); err != nil {
+		return false
+	}
+	return other.Membership == event.MembershipJoin || other.Membership == event.MembershipInvite
+}
+
+// publishDirect appends the room to the bot's m.direct account data, so the
+// room is labelled as a DM rather than a nameless private room.
+func (b *Base) publishDirect(userID id.UserID, roomID id.RoomID) {
+	ctx := context.Background()
+
+	dmRooms := map[id.UserID][]id.RoomID{}
+	// A missing m.direct is a 404 — start from empty rather than bailing.
+	_ = b.Client.GetAccountData(ctx, "m.direct", &dmRooms)
+
+	for _, existing := range dmRooms[userID] {
+		if existing == roomID {
+			return
+		}
+	}
+	dmRooms[userID] = append(dmRooms[userID], roomID)
+
+	if err := b.Client.SetAccountData(ctx, "m.direct", dmRooms); err != nil {
+		slog.Warn("publish m.direct", "user", userID, "room", roomID, "err", err)
+	}
+}
+
+// RecordDMRoom claims a room as the user's DM room. Used for user-initiated
+// DM invites, where the invite itself is the intent — it overwrites any older
+// mapping, since the user just told us which room they want to talk in.
+func (b *Base) RecordDMRoom(userID id.UserID, roomID id.RoomID) {
+	slog.Info("recorded user-initiated DM room", "user", userID, "room", roomID)
+	rememberDMRoom(userID, roomID)
+	b.publishDirect(userID, roomID)
+}
+
+// LearnDMRoom records a room the user messaged the bot in as their DM room,
+// when it really is a two-person room. This adopts DM rooms that predate the
+// database mapping instead of leaving them orphaned beside a freshly created one.
+func (b *Base) LearnDMRoom(userID id.UserID, roomID id.RoomID) {
+	if b == nil || b.Client == nil {
+		return
+	}
+
+	dmCacheMu.Lock()
+	_, known := dmCache[userID]
+	mapped := dmMapped[userID]
+	notDM := notDMCache[roomID]
+	dmCacheMu.Unlock()
+	if known || mapped || notDM {
+		return
+	}
+	if _, ok := storedDMRoom(userID); ok {
+		// Note it as mapped, not as resolved: caching the room here would let
+		// GetDMRoom skip its liveness check on a possibly-stale room.
+		dmCacheMu.Lock()
+		dmMapped[userID] = true
+		dmCacheMu.Unlock()
+		return
+	}
+
+	members, err := b.Client.JoinedMembers(context.Background(), roomID)
+	if err != nil {
+		return
+	}
+	_, botIn := members.Joined[b.Client.UserID]
+	_, userIn := members.Joined[userID]
+	if len(members.Joined) != 2 || !botIn || !userIn {
+		// Group room — remember that so every later message here is free.
+		dmCacheMu.Lock()
+		notDMCache[roomID] = true
+		dmCacheMu.Unlock()
+		return
+	}
+
+	slog.Info("adopted existing DM room", "user", userID, "room", roomID)
+	rememberDMRoom(userID, roomID)
+}
+
+// findExistingDMRoom scans the bot's joined rooms for a two-person room shared
+// with userID. The scan is expensive (one member lookup per room), so it runs
+// at most once per process and only on the path that would otherwise create a
+// duplicate room.
+func (b *Base) findExistingDMRoom(userID id.UserID) (id.RoomID, bool) {
+	dmSweepOnce.Do(func() {
+		ctx := context.Background()
+		joined, err := b.Client.JoinedRooms(ctx)
+		if err != nil {
+			slog.Warn("DM sweep: list joined rooms", "err", err)
+			return
+		}
+		for _, roomID := range joined.JoinedRooms {
+			members, err := b.Client.JoinedMembers(ctx, roomID)
+			if err != nil || len(members.Joined) != 2 {
+				continue
+			}
+			if _, ok := members.Joined[b.Client.UserID]; !ok {
+				continue
+			}
+			for member := range members.Joined {
+				if member != b.Client.UserID {
+					dmSweepIndex[member] = roomID
+				}
+			}
+		}
+		slog.Info("DM sweep complete", "rooms", len(joined.JoinedRooms), "dms", len(dmSweepIndex))
+	})
+
+	roomID, ok := dmSweepIndex[userID]
+	return roomID, ok
+}
+
 // GetDMRoom returns the DM room for a user, creating one if needed.
 func (b *Base) GetDMRoom(userID id.UserID) (id.RoomID, error) {
 	dmCacheMu.Lock()
@@ -695,17 +877,42 @@ func (b *Base) GetDMRoom(userID id.UserID) (id.RoomID, error) {
 	}
 	dmCacheMu.Unlock()
 
-	// Check account data for existing DM rooms
-	var dmRooms map[id.UserID][]id.RoomID
-	err := b.Client.GetAccountData(context.Background(), "m.direct", &dmRooms)
-	if err == nil {
-		if rooms, ok := dmRooms[userID]; ok && len(rooms) > 0 {
-			roomID := rooms[len(rooms)-1] // use most recent
+	// Persisted mapping — the authoritative store across restarts.
+	if roomID, ok := storedDMRoom(userID); ok {
+		if b.dmRoomUsable(roomID, userID) {
 			dmCacheMu.Lock()
 			dmCache[userID] = roomID
 			dmCacheMu.Unlock()
 			return roomID, nil
 		}
+		slog.Info("stored DM room no longer usable, recreating", "user", userID, "room", roomID)
+		forgetDMRoom(userID)
+	}
+
+	// Check account data for existing DM rooms
+	var dmRooms map[id.UserID][]id.RoomID
+	err := b.Client.GetAccountData(context.Background(), "m.direct", &dmRooms)
+	if err == nil {
+		for i := len(dmRooms[userID]) - 1; i >= 0; i-- {
+			roomID := dmRooms[userID][i] // most recent first
+			if b.dmRoomUsable(roomID, userID) {
+				rememberDMRoom(userID, roomID)
+				return roomID, nil
+			}
+		}
+	}
+
+	// Last resort before creating: sweep the rooms the bot is already in for a
+	// two-person room shared with this user. Users who predate the dm_rooms
+	// table have such a room and nothing pointing at it — without this sweep
+	// their first post-upgrade DM would open yet another duplicate. If several
+	// duplicates exist we cannot tell which is liveliest (no /sync, so no
+	// timestamps), so the newest by room-list order wins.
+	if roomID, ok := b.findExistingDMRoom(userID); ok {
+		slog.Info("recovered pre-existing DM room by member sweep", "user", userID, "room", roomID)
+		rememberDMRoom(userID, roomID)
+		b.publishDirect(userID, roomID)
+		return roomID, nil
 	}
 
 	// No existing DM room — create one
@@ -728,9 +935,8 @@ func (b *Base) GetDMRoom(userID id.UserID) (id.RoomID, error) {
 		return "", fmt.Errorf("create DM room: %w", err)
 	}
 
-	dmCacheMu.Lock()
-	dmCache[userID] = resp.RoomID
-	dmCacheMu.Unlock()
+	rememberDMRoom(userID, resp.RoomID)
+	b.publishDirect(userID, resp.RoomID)
 	return resp.RoomID, nil
 }
 

internal/plugin/tarot.go

@@ -4,7 +4,6 @@ import (
 	"fmt"
 	"log/slog"
 	"math/rand"
-	"os"
 	"strings"
 
 	"gogobee/internal/db"
@@ -116,9 +115,7 @@ func (p *TarotPlugin) handleTarot(ctx MessageContext) error {
 		return p.SendReply(ctx.RoomID, ctx.EventID, "You've used up your readings for today. The cards need rest, even if you don't.")
 	}
 
-	ollamaHost := os.Getenv("OLLAMA_HOST")
-	ollamaModel := os.Getenv("OLLAMA_MODEL")
-	if ollamaHost == "" || ollamaModel == "" {
+	if !llmConfigured() {
 		return p.SendReply(ctx.RoomID, ctx.EventID, "Tarot reader is on a union-mandated vacation and will return when morale improves.")
 	}
 
@@ -151,7 +148,7 @@ func (p *TarotPlugin) handleTarot(ctx MessageContext) error {
 	}
 	prompt := fmt.Sprintf("%s\n%s\n\n%s\n\nCard drawn: %s%s\nGive the reading.", tarotBasePrompt, tarotSingleSuffix, tarotFewShot, card, extraLines)
 
-	response, err := callOllama(ollamaHost, ollamaModel, prompt)
+	response, err := callLLM(prompt)
 	if err != nil {
 		slog.Error("tarot: ollama call", "err", err)
 		return p.SendReply(ctx.RoomID, ctx.EventID, "Tarot reader is on a union-mandated vacation and will return when morale improves.")
@@ -166,9 +163,7 @@ func (p *TarotPlugin) handleSpread(ctx MessageContext) error {
 		return p.SendReply(ctx.RoomID, ctx.EventID, "You've used up your readings for today. The cards need rest, even if you don't.")
 	}
 
-	ollamaHost := os.Getenv("OLLAMA_HOST")
-	ollamaModel := os.Getenv("OLLAMA_MODEL")
-	if ollamaHost == "" || ollamaModel == "" {
+	if !llmConfigured() {
 		return p.SendReply(ctx.RoomID, ctx.EventID, "Tarot reader is on a union-mandated vacation and will return when morale improves.")
 	}
 
@@ -202,7 +197,7 @@ func (p *TarotPlugin) handleSpread(ctx MessageContext) error {
 	prompt := fmt.Sprintf("%s\n%s\n\n%s\n\nCards drawn:\n- Past: %s\n- Present: %s\n- Future: %s%s\nGive the reading.",
 		tarotBasePrompt, tarotSpreadSuffix, tarotFewShot, cards[0], cards[1], cards[2], extraLines)
 
-	response, err := callOllama(ollamaHost, ollamaModel, prompt)
+	response, err := callLLM(prompt)
 	if err != nil {
 		slog.Error("tarot: ollama call", "err", err)
 		return p.SendReply(ctx.RoomID, ctx.EventID, "Tarot reader is on a union-mandated vacation and will return when morale improves.")

internal/plugin/vibe.go

@@ -120,9 +120,7 @@ func (p *VibePlugin) resetCooldown(roomID id.RoomID) {
 }
 
 func (p *VibePlugin) handleVibe(ctx MessageContext) error {
-	ollamaHost := os.Getenv("OLLAMA_HOST")
-	ollamaModel := os.Getenv("OLLAMA_MODEL")
-	if ollamaHost == "" || ollamaModel == "" {
+	if !llmConfigured() {
 		return p.SendReply(ctx.RoomID, ctx.EventID, "LLM is not configured.")
 	}
 
@@ -154,7 +152,7 @@ Describe the room's current vibe:`, botName, transcript)
 		slog.Error("vibe: send thinking", "err", err)
 	}
 
-	response, err := callOllama(ollamaHost, ollamaModel, prompt)
+	response, err := callLLM(prompt)
 	if err != nil {
 		slog.Error("vibe: ollama call", "err", err)
 		p.resetCooldown(ctx.RoomID) // Don't consume cooldown on failure
@@ -165,9 +163,7 @@ Describe the room's current vibe:`, botName, transcript)
 }
 
 func (p *VibePlugin) handleTLDR(ctx MessageContext) error {
-	ollamaHost := os.Getenv("OLLAMA_HOST")
-	ollamaModel := os.Getenv("OLLAMA_MODEL")
-	if ollamaHost == "" || ollamaModel == "" {
+	if !llmConfigured() {
 		return p.SendReply(ctx.RoomID, ctx.EventID, "LLM is not configured.")
 	}
 
@@ -199,7 +195,7 @@ Summary:`, tldrBotName, transcript)
 		slog.Error("vibe: send thinking", "err", err)
 	}
 
-	response, err := callOllama(ollamaHost, ollamaModel, prompt)
+	response, err := callLLM(prompt)
 	if err != nil {
 		slog.Error("vibe: ollama call", "err", err)
 		p.resetCooldown(ctx.RoomID) // Don't consume cooldown on failure

internal/plugin/wotd.go

@@ -1,19 +1,17 @@
 package plugin
 
 import (
-	"bytes"
+	"context"
 	"database/sql"
 	"encoding/json"
 	"fmt"
-	"io"
 	"log/slog"
-	"net/http"
-	"os"
 	"strings"
 	"time"
 
 	"gogobee/internal/db"
 	"gogobee/internal/dreamclient"
+	"gogobee/internal/llm"
 
 	"maunium.net/go/mautrix"
 	"maunium.net/go/mautrix/id"
@@ -533,9 +531,7 @@ func (p *WOTDPlugin) trackUsage(ctx MessageContext) {
 // verifyUsage asks the LLM whether the word was used correctly in context.
 // Returns false if LLM is not configured or on any error.
 func (p *WOTDPlugin) verifyUsage(word, message string) bool {
-	host := os.Getenv("OLLAMA_HOST")
-	model := os.Getenv("OLLAMA_MODEL")
-	if host == "" || model == "" {
+	if !llmConfigured() {
 		return false
 	}
 
@@ -545,58 +541,30 @@ The Word of the Day is "%s". Was this word used correctly and meaningfully in th
 
 Respond with ONLY "yes" or "no".`, message, word)
 
-	payload := map[string]interface{}{
-		"model":  model,
-		"prompt": prompt,
-		"stream": false,
-		"think":  false,
-	}
-	data, err := json.Marshal(payload)
-	if err != nil {
-		return false
-	}
-
-	apiURL := strings.TrimRight(host, "/") + "/api/generate"
-	slog.Debug("wotd: sending LLM verification request", "url", apiURL, "word", word)
-	client := &http.Client{Timeout: 30 * time.Second}
-	resp, err := client.Post(apiURL, "application/json", bytes.NewReader(data))
+	response, err := llmGenerate(context.Background(), llm.Request{
+		Prompt:  prompt,
+		Timeout: wotdLLMTimeout,
+	})
 	if err != nil {
 		slog.Error("wotd: LLM verify request failed", "err", err)
 		return false
 	}
-	defer resp.Body.Close()
 
-	body, err := io.ReadAll(resp.Body)
-	if err != nil || resp.StatusCode != http.StatusOK {
-		return false
-	}
-
-	var result struct {
-		Response string `json:"response"`
-	}
-	if json.Unmarshal(body, &result) != nil {
-		return false
-	}
-
-	response := result.Response
-	// Strip <think>...</think> blocks (Qwen 3.5 reasoning)
-	if i := strings.Index(response, "<think>"); i != -1 {
-		if j := strings.Index(response, "</think>"); j != -1 {
-			response = response[:i] + response[j+len("</think>"):]
-		}
-	}
 	answer := strings.ToLower(strings.TrimSpace(response))
 	accepted := strings.HasPrefix(answer, "yes")
 	slog.Debug("wotd: LLM verification", "word", word, "answer", answer, "accepted", accepted)
 	return accepted
 }
 
+// wotdLLMTimeout preserves the 30s cap both WOTD paths enforced with their own
+// http.Client. These run on message traffic, so they stay well under the
+// interactive default.
+const wotdLLMTimeout = 30 * time.Second
+
 // llmTranslate asks the LLM for a brief English translation of a foreign word.
 // Returns empty string on failure.
 func (p *WOTDPlugin) llmTranslate(word, lang string) string {
-	host := os.Getenv("OLLAMA_HOST")
-	model := os.Getenv("OLLAMA_MODEL")
-	if host == "" || model == "" {
+	if !llmConfigured() {
 		return ""
 	}
 
@@ -612,44 +580,15 @@ func (p *WOTDPlugin) llmTranslate(word, lang string) string {
 		`Translate the %s word "%s" into English. Reply with ONLY the English translation — one or two words, no explanation, no punctuation.`,
 		langName, word)
 
-	payload := map[string]interface{}{
-		"model":  model,
-		"prompt": prompt,
-		"stream": false,
-		"think":  false,
-	}
-	data, err := json.Marshal(payload)
-	if err != nil {
-		return ""
-	}
-
-	apiURL := strings.TrimRight(host, "/") + "/api/generate"
-	client := &http.Client{Timeout: 30 * time.Second}
-	resp, err := client.Post(apiURL, "application/json", bytes.NewReader(data))
+	response, err := llmGenerate(context.Background(), llm.Request{
+		Prompt:  prompt,
+		Timeout: wotdLLMTimeout,
+	})
 	if err != nil {
 		slog.Error("wotd: LLM translate request failed", "err", err)
 		return ""
 	}
-	defer resp.Body.Close()
 
-	body, err := io.ReadAll(resp.Body)
-	if err != nil || resp.StatusCode != http.StatusOK {
-		return ""
-	}
-
-	var result struct {
-		Response string `json:"response"`
-	}
-	if json.Unmarshal(body, &result) != nil {
-		return ""
-	}
-
-	response := result.Response
-	if i := strings.Index(response, "<think>"); i != -1 {
-		if j := strings.Index(response, "</think>"); j != -1 {
-			response = response[:i] + response[j+len("</think>"):]
-		}
-	}
 	translation := strings.TrimSpace(response)
 	if translation == "" || len(translation) > 50 {
 		return ""

main.go

@@ -13,6 +13,7 @@ import (
 	"gogobee/internal/bot"
 	"gogobee/internal/db"
 	"gogobee/internal/dreamclient"
+	"gogobee/internal/llm"
 	"gogobee/internal/peteclient"
 	"gogobee/internal/plugin"
 	"gogobee/internal/util"
@@ -37,9 +38,11 @@ func main() {
 		logLevel = "info"
 	}
 	util.InitLogger(logLevel)
+	llmCfg := llm.ConfigFromEnv()
 	slog.Info(version.Full(), "level", logLevel,
-		"ollama_host", os.Getenv("OLLAMA_HOST"),
-		"ollama_model", os.Getenv("OLLAMA_MODEL"))
+		"llm_backend", llmCfg.Backend,
+		"llm_endpoint", llmCfg.Endpoint,
+		"llm_model", llmCfg.Model)
 
 	dataDir := os.Getenv("DATA_DIR")
 	if dataDir == "" {
@@ -252,6 +255,9 @@ func main() {
 
 	// ---- Set up event handlers ----
 
+	// Minimal Base used only for DM-room bookkeeping from the event handlers.
+	dmLearner := &plugin.Base{Client: client}
+
 	// Auto-join on invite + moderation member tracking
 	sess.OnEventType(event.StateMember, func(ctx context.Context, evt *event.Event) {
 		defer func() {
@@ -274,6 +280,11 @@ func main() {
 					slog.Error("failed to join room", "room", evt.RoomID, "err", err)
 				} else {
 					slog.Info("joined room", "room", evt.RoomID)
+					// A user-initiated DM invite: claim it now, before any
+					// outbound DM has a chance to create a rival room.
+					if mem.IsDirect {
+						dmLearner.RecordDMRoom(evt.Sender, evt.RoomID)
+					}
 				}
 			}
 			return
@@ -322,6 +333,10 @@ func main() {
 				body = strings.TrimSpace(body[idx+2:])
 			}
 		}
+		// Adopt the room the user is talking in if it's their DM room, so the
+		// bot replies there instead of opening a second one.
+		dmLearner.LearnDMRoom(evt.Sender, evt.RoomID)
+
 		msgCtx := plugin.MessageContext{
 			RoomID:    evt.RoomID,
 			EventID:   evt.ID,